Tday Demo Video on X

Tday Screenshot 1 Tday Screenshot 2

Tday Screenshot 3 Tday Screenshot 4

Tday Screenshot 5 Tday Screenshot 6

Image

🧑‍💻 User Experience

🔥Computer Use (DEMO: https://x.com/unbug/status/2053205496421921278?s=20)
Give any agent hands on your desktop — no vision model required. Uses native Accessibility APIs to read and control any app directly, without screenshots. One-click enable for Claude Code, Codex, Gemini, OpenCode, and Pi. Agents can click UI elements, type, scroll, open apps, control Chrome via CDP, and fall back to OCR when needed. Built-in Computer Operator CoWorker + one-click permission setup. Runs natively on macOS (arm64 + x64), Windows x64, and Linux x64.

1. Multi-tab workspace with drag-reorder
Open unlimited agent tabs simultaneously, drag to reorder, wrap to multiple rows. Per-tab working directory with staged commit (Enter / Apply / Browse) and last-cwd persistence across restarts.

2. Unified agent history — browse & restore every conversation
Every session across all 9 agents is automatically captured and indexed. Browse the full history from the Logo menu, search by content, filter by agent, and restore any past conversation with one click — including the exact working directory and agent session ID, so the conversation resumes exactly where it left off.

3. Cross-agent usage analytics — tokens, cost, every session
Complete token and estimated cost tracking for every agent across every session. Dashboard shows total spend summary cards, a daily bar chart, a by-model breakdown table, and a by-agent breakdown table — all backed by a local SQLite store with 30+ model pricing. Filter by time period (today / 7d / 30d / 90d / custom) or by agent. Nothing leaves your machine.

4. 9 AI agent adapters
First-class support for Pi, Claude Code, Codex CLI, GitHub Copilot CLI, OpenCode, Gemini CLI, Qwen-Code, Crush, and Hermes — each with auto-detect, install, and per-agent accent color.

5. CoWorker system — role personas as system prompts
Reusable specialist personas injected as system prompts. Three source types: built-in (AGENT.md files), online (URL-cached from GitHub), custom (inline text / local file / URL). Includes 8 built-in engineer roles and 3 curated presets: Karpathy Code Guidelines, Devin-style Planner, Earnings Call Analyst.

6. Per-tab CoWorker selector
CoWorker dropdown in the CwdBar next to the working-directory input. Selecting a CoWorker immediately writes its system prompt into the running PTY — no tab restart required.

7. Cron job scheduler
Settings → Cron: schedule automated agent tasks with Interval / At Time / Custom cron modes, human-readable preview, enable/disable, clone, and delete. Dashboard shows next-run countdown, run count, and last-status per job.

8. Agent-native session resume
Conversation continuity across restarts: Claude Code (--resume), Codex (resume <id>), OpenCode (--session). Session ID is captured from agent output and restored on tab reopen.

9. Tab session persistence
All open tabs (id, title, agent, cwd) are saved on every change and fully restored on next launch, with the last-active tab focused first.

10. "Keep Awake" power management
One-click toggle in the Logo menu to block system sleep while long-running agent jobs are in progress — without affecting screen dimming.

11. In-app update checker
Polls GitHub Releases API 10 s after launch, then every 30 min. Green dot badge when a new version is available, with links to release notes and download.

⚙️ Configuration & Integration

12. 28-provider settings panel
CRUD UI for provider profiles covering DeepSeek, OpenAI, Anthropic, Google Gemini, xAI, Groq, Mistral, Ollama, LM Studio, OpenRouter, and 18 more — with OpenAI-compatible and Anthropic-compatible base URL modes.

13. Per-agent provider & model binding
Each agent tab can use a different provider and model. Model flags are projected to CLI arguments automatically at spawn time. "Use one provider for all agents" shared-config toggle also available.

14. Local inference auto-discovery
TCP pre-filter + HTTP fingerprint scan for Ollama, LM Studio, vLLM, llama.cpp, SGLang, LocalAI, and Jan. Discovered models appear as chips in the provider UI. Manual base-URL probe also available.

15. Reliable initialPrompt delivery
Task prompts reach the agent at spawn time even when the screen is locked. CLI positional arg for Codex / Claude Code / OpenCode / Gemini / Qwen-Code; bracketed-paste PTY write (ESC[200~) for Pi / Copilot / Crush / Hermes.

🔧 Technical Foundation

16. Real PTY terminal with xterm.js
Full bidirectional PTY via node-pty + xterm.js with fit, resize, and web-links addons. Correct COLUMNS/LINES env vars and RAF-timed spawn ensure accurate terminal dimensions from the first keystroke.

17. DeepSeek Anthropic gateway proxy
In-process HTTP server translating OpenAI Responses API → Anthropic Messages API — enabling streaming, extended thinking, tool use, and multi-turn conversations with DeepSeek models through Claude Code.

18. Resizable settings dialog
Drag handle in the bottom-right corner. Dialog size is lazy-initialized to fit the current viewport, so the tab row is never pushed out of the visible area.

19. Electron drag-region safety
All interactive elements (tab buttons, dropdowns, settings tabs, backdrop) are marked no-drag (-webkit-app-region: no-drag), preventing Electron's window-drag layer from swallowing clicks — especially when the TabBar wraps to multiple rows.

20. Windows & cross-platform support
PATH augmentation for nvm-windows, Volta, fnm, and other version managers. Native menu preserved on Windows. Consistent PTY column/row handling and build pipeline across macOS (arm64 + x64), Windows, and Linux.

🧑‍💻 User Experience

🔥Computer Use (DEMO: https://x.com/unbug/status/2053205496421921278?s=20)
Give any agent hands on your desktop — no vision model required. Uses native Accessibility APIs to read and control any app directly, without screenshots. One-click enable for Claude Code, Codex, Gemini, OpenCode, and Pi. Agents can click UI elements, type, scroll, open apps, control Chrome via CDP, and fall back to OCR when needed. Built-in Computer Operator CoWorker + one-click permission setup. Runs natively on macOS (arm64 + x64), Windows x64, and Linux x64.

1. Multi-tab workspace with drag-reorder
Open unlimited agent tabs simultaneously, drag to reorder, wrap to multiple rows. Per-tab working directory with staged commit (Enter / Apply / Browse) and last-cwd persistence across restarts.

2. Unified agent history — browse & restore every conversation
Every session across all 9 agents is automatically captured and indexed. Browse the full history from the Logo menu, search by content, filter by agent, and restore any past conversation with one click — including the exact working directory and agent session ID, so the conversation resumes exactly where it left off.

3. Cross-agent usage analytics — tokens, cost, every session
Complete token and estimated cost tracking for every agent across every session. Dashboard shows total spend summary cards, a daily bar chart, a by-model breakdown table, and a by-agent breakdown table — all backed by a local SQLite store with 30+ model pricing. Filter by time period (today / 7d / 30d / 90d / custom) or by agent. Nothing leaves your machine.

4. 9 AI agent adapters
First-class support for Pi, Claude Code, Codex CLI, GitHub Copilot CLI, OpenCode, Gemini CLI, Qwen-Code, Crush, and Hermes — each with auto-detect, install, and per-agent accent color.

5. CoWorker system — role personas as system prompts
Reusable specialist personas injected as system prompts. Three source types: built-in (AGENT.md files), online (URL-cached from GitHub), custom (inline text / local file / URL). Includes 8 built-in engineer roles and 3 curated presets: Karpathy Code Guidelines, Devin-style Planner, Earnings Call Analyst.

6. Per-tab CoWorker selector
CoWorker dropdown in the CwdBar next to the working-directory input. Selecting a CoWorker immediately writes its system prompt into the running PTY — no tab restart required.

7. Cron job scheduler
Settings → Cron: schedule automated agent tasks with Interval / At Time / Custom cron modes, human-readable preview, enable/disable, clone, and delete. Dashboard shows next-run countdown, run count, and last-status per job.

8. Agent-native session resume
Conversation continuity across restarts: Claude Code (--resume), Codex (resume <id>), OpenCode (--session). Session ID is captured from agent output and restored on tab reopen.

9. Tab session persistence
All open tabs (id, title, agent, cwd) are saved on every change and fully restored on next launch, with the last-active tab focused first.

10. "Keep Awake" power management
One-click toggle in the Logo menu to block system sleep while long-running agent jobs are in progress — without affecting screen dimming.

11. In-app update checker
Polls GitHub Releases API 10 s after launch, then every 30 min. Green dot badge when a new version is available, with links to release notes and download.

⚙️ Configuration & Integration

12. 28-provider settings panel
CRUD UI for provider profiles covering DeepSeek, OpenAI, Anthropic, Google Gemini, xAI, Groq, Mistral, Ollama, LM Studio, OpenRouter, and 18 more — with OpenAI-compatible and Anthropic-compatible base URL modes.

13. Per-agent provider & model binding
Each agent tab can use a different provider and model. Model flags are projected to CLI arguments automatically at spawn time. "Use one provider for all agents" shared-config toggle also available.

14. Local inference auto-discovery
TCP pre-filter + HTTP fingerprint scan for Ollama, LM Studio, vLLM, llama.cpp, SGLang, LocalAI, and Jan. Discovered models appear as chips in the provider UI. Manual base-URL probe also available.

15. Reliable initialPrompt delivery
Task prompts reach the agent at spawn time even when the screen is locked. CLI positional arg for Codex / Claude Code / OpenCode / Gemini / Qwen-Code; bracketed-paste PTY write (ESC[200~) for Pi / Copilot / Crush / Hermes.

🔧 Technical Foundation

16. Real PTY terminal with xterm.js
Full bidirectional PTY via node-pty + xterm.js with fit, resize, and web-links addons. Correct COLUMNS/LINES env vars and RAF-timed spawn ensure accurate terminal dimensions from the first keystroke.

17. DeepSeek Anthropic gateway proxy
In-process HTTP server translating OpenAI Responses API → Anthropic Messages API — enabling streaming, extended thinking, tool use, and multi-turn conversations with DeepSeek models through Claude Code.

18. Resizable settings dialog
Drag handle in the bottom-right corner. Dialog size is lazy-initialized to fit the current viewport, so the tab row is never pushed out of the visible area.

19. Electron drag-region safety
All interactive elements (tab buttons, dropdowns, settings tabs, backdrop) are marked no-drag (-webkit-app-region: no-drag), preventing Electron's window-drag layer from swallowing clicks — especially when the TabBar wraps to multiple rows.

20. Windows & cross-platform support
PATH augmentation for nvm-windows, Volta, fnm, and other version managers. Native menu preserved on Windows. Consistent PTY column/row handling and build pipeline across macOS (arm64 + x64), Windows, and Linux.

🧑‍💻 User Experience

🔥Computer Use (DEMO: https://x.com/unbug/status/2053205496421921278?s=20)
Give any agent hands on your desktop — no vision model required. Uses native Accessibility APIs to read and control any app directly, without screenshots. One-click enable for Claude Code, Codex, Gemini, OpenCode, and Pi. Agents can click UI elements, type, scroll, open apps, control Chrome via CDP, and fall back to OCR when needed. Built-in Computer Operator CoWorker + one-click permission setup. Runs natively on macOS (arm64 + x64), Windows x64, and Linux x64.

1. Multi-tab workspace with drag-reorder
Open unlimited agent tabs simultaneously, drag to reorder, wrap to multiple rows. Per-tab working directory with staged commit (Enter / Apply / Browse) and last-cwd persistence across restarts.

2. Unified agent history — browse & restore every conversation
Every session across all 9 agents is automatically captured and indexed. Browse the full history from the Logo menu, search by content, filter by agent, and restore any past conversation with one click — including the exact working directory and agent session ID, so the conversation resumes exactly where it left off.

3. Cross-agent usage analytics — tokens, cost, every session
Complete token and estimated cost tracking for every agent across every session. Dashboard shows total spend summary cards, a daily bar chart, a by-model breakdown table, and a by-agent breakdown table — all backed by a local SQLite store with 30+ model pricing. Filter by time period (today / 7d / 30d / 90d / custom) or by agent. Nothing leaves your machine.

4. 9 AI agent adapters
First-class support for Pi, Claude Code, Codex CLI, GitHub Copilot CLI, OpenCode, Gemini CLI, Qwen-Code, Crush, and Hermes — each with auto-detect, install, and per-agent accent color.

5. CoWorker system — role personas as system prompts
Reusable specialist personas injected as system prompts. Three source types: built-in (AGENT.md files), online (URL-cached from GitHub), custom (inline text / local file / URL). Includes 8 built-in engineer roles and 3 curated presets: Karpathy Code Guidelines, Devin-style Planner, Earnings Call Analyst.

6. Per-tab CoWorker selector
CoWorker dropdown in the CwdBar next to the working-directory input. Selecting a CoWorker immediately writes its system prompt into the running PTY — no tab restart required.

7. Cron job scheduler
Settings → Cron: schedule automated agent tasks with Interval / At Time / Custom cron modes, human-readable preview, enable/disable, clone, and delete. Dashboard shows next-run countdown, run count, and last-status per job.

8. Agent-native session resume
Conversation continuity across restarts: Claude Code (--resume), Codex (resume <id>), OpenCode (--session). Session ID is captured from agent output and restored on tab reopen.

9. Tab session persistence
All open tabs (id, title, agent, cwd) are saved on every change and fully restored on next launch, with the last-active tab focused first.

10. "Keep Awake" power management
One-click toggle in the Logo menu to block system sleep while long-running agent jobs are in progress — without affecting screen dimming.

11. In-app update checker
Polls GitHub Releases API 10 s after launch, then every 30 min. Green dot badge when a new version is available, with links to release notes and download.

⚙️ Configuration & Integration

12. 28-provider settings panel
CRUD UI for provider profiles covering DeepSeek, OpenAI, Anthropic, Google Gemini, xAI, Groq, Mistral, Ollama, LM Studio, OpenRouter, and 18 more — with OpenAI-compatible and Anthropic-compatible base URL modes.

13. Per-agent provider & model binding
Each agent tab can use a different provider and model. Model flags are projected to CLI arguments automatically at spawn time. "Use one provider for all agents" shared-config toggle also available.

14. Local inference auto-discovery
TCP pre-filter + HTTP fingerprint scan for Ollama, LM Studio, vLLM, llama.cpp, SGLang, LocalAI, and Jan. Discovered models appear as chips in the provider UI. Manual base-URL probe also available.

15. Reliable initialPrompt delivery
Task prompts reach the agent at spawn time even when the screen is locked. CLI positional arg for Codex / Claude Code / OpenCode / Gemini / Qwen-Code; bracketed-paste PTY write (ESC[200~) for Pi / Copilot / Crush / Hermes.

🔧 Technical Foundation

16. Real PTY terminal with xterm.js
Full bidirectional PTY via node-pty + xterm.js with fit, resize, and web-links addons. Correct COLUMNS/LINES env vars and RAF-timed spawn ensure accurate terminal dimensions from the first keystroke.

17. DeepSeek Anthropic gateway proxy
In-process HTTP server translating OpenAI Responses API → Anthropic Messages API — enabling streaming, extended thinking, tool use, and multi-turn conversations with DeepSeek models through Claude Code.

18. Resizable settings dialog
Drag handle in the bottom-right corner. Dialog size is lazy-initialized to fit the current viewport, so the tab row is never pushed out of the visible area.

19. Electron drag-region safety
All interactive elements (tab buttons, dropdowns, settings tabs, backdrop) are marked no-drag (-webkit-app-region: no-drag), preventing Electron's window-drag layer from swallowing clicks — especially when the TabBar wraps to multiple rows.

20. Windows & cross-platform support
PATH augmentation for nvm-windows, Volta, fnm, and other version managers. Native menu preserved on Windows. Consistent PTY column/row handling and build pipeline across macOS (arm64 + x64), Windows, and Linux.

latest

🧑‍💻 User Experience

🔥Computer Use
Give any agent hands on your desktop — no vision model required. Uses native Accessibility APIs to read and control any app directly, without screenshots. One-click enable for Claude Code, Codex, Gemini, OpenCode, and Pi. Agents can click UI elements, type, scroll, open apps, control Chrome via CDP, and fall back to OCR when needed. Built-in Computer Operator CoWorker + one-click permission setup. Runs natively on macOS (arm64 + x64), Windows x64, and Linux x64.

1. Multi-tab workspace with drag-reorder
Open unlimited agent tabs simultaneously, drag to reorder, wrap to multiple rows. Per-tab working directory with staged commit (Enter / Apply / Browse) and last-cwd persistence across restarts.

2. Unified agent history — browse & restore every conversation
Every session across all 9 agents is automatically captured and indexed. Browse the full history from the Logo menu, search by content, filter by agent, and restore any past conversation with one click — including the exact working directory and agent session ID, so the conversation resumes exactly where it left off.

3. Cross-agent usage analytics — tokens, cost, every session
Complete token and estimated cost tracking for every agent across every session. Dashboard shows total spend summary cards, a daily bar chart, a by-model breakdown table, and a by-agent breakdown table — all backed by a local SQLite store with 30+ model pricing. Filter by time period (today / 7d / 30d / 90d / custom) or by agent. Nothing leaves your machine.

4. 9 AI agent adapters
First-class support for Pi, Claude Code, Codex CLI, GitHub Copilot CLI, OpenCode, Gemini CLI, Qwen-Code, Crush, and Hermes — each with auto-detect, install, and per-agent accent color.

5. CoWorker system — role personas as system prompts
Reusable specialist personas injected as system prompts. Three source types: built-in (AGENT.md files), online (URL-cached from GitHub), custom (inline text / local file / URL). Includes 8 built-in engineer roles and 3 curated presets: Karpathy Code Guidelines, Devin-style Planner, Earnings Call Analyst.

6. Per-tab CoWorker selector
CoWorker dropdown in the CwdBar next to the working-directory input. Selecting a CoWorker immediately writes its system prompt into the running PTY — no tab restart required.

7. Cron job scheduler
Settings → Cron: schedule automated agent tasks with Interval / At Time / Custom cron modes, human-readable preview, enable/disable, clone, and delete. Dashboard shows next-run countdown, run count, and last-status per job.

8. Agent-native session resume
Conversation continuity across restarts: Claude Code (--resume), Codex (resume <id>), OpenCode (--session). Session ID is captured from agent output and restored on tab reopen.

9. Tab session persistence
All open tabs (id, title, agent, cwd) are saved on every change and fully restored on next launch, with the last-active tab focused first.

10. "Keep Awake" power management
One-click toggle in the Logo menu to block system sleep while long-running agent jobs are in progress — without affecting screen dimming.

11. In-app update checker
Polls GitHub Releases API 10 s after launch, then every 30 min. Green dot badge when a new version is available, with links to release notes and download.

⚙️ Configuration & Integration

12. 28-provider settings panel
CRUD UI for provider profiles covering DeepSeek, OpenAI, Anthropic, Google Gemini, xAI, Groq, Mistral, Ollama, LM Studio, OpenRouter, and 18 more — with OpenAI-compatible and Anthropic-compatible base URL modes.

13. Per-agent provider & model binding
Each agent tab can use a different provider and model. Model flags are projected to CLI arguments automatically at spawn time. "Use one provider for all agents" shared-config toggle also available.

14. Local inference auto-discovery
TCP pre-filter + HTTP fingerprint scan for Ollama, LM Studio, vLLM, llama.cpp, SGLang, LocalAI, and Jan. Discovered models appear as chips in the provider UI. Manual base-URL probe also available.

15. Reliable initialPrompt delivery
Task prompts reach the agent at spawn time even when the screen is locked. CLI positional arg for Codex / Claude Code / OpenCode / Gemini / Qwen-Code; bracketed-paste PTY write (ESC[200~) for Pi / Copilot / Crush / Hermes.

🔧 Technical Foundation

16. Real PTY terminal with xterm.js
Full bidirectional PTY via node-pty + xterm.js with fit, resize, and web-links addons. Correct COLUMNS/LINES env vars and RAF-timed spawn ensure accurate terminal dimensions from the first keystroke.

17. DeepSeek Anthropic gateway proxy
In-process HTTP server translating OpenAI Responses API → Anthropic Messages API — enabling streaming, extended thinking, tool use, and multi-turn conversations with DeepSeek models through Claude Code.

18. Resizable settings dialog
Drag handle in the bottom-right corner. Dialog size is lazy-initialized to fit the current viewport, so the tab row is never pushed out of the visible area.

19. Electron drag-region safety
All interactive elements (tab buttons, dropdowns, settings tabs, backdrop) are marked no-drag (-webkit-app-region: no-drag), preventing Electron's window-drag layer from swallowing clicks — especially when the TabBar wraps to multiple rows.

20. Windows & cross-platform support
PATH augmentation for nvm-windows, Volta, fnm, and other version managers. Native menu preserved on Windows. Consistent PTY column/row handling and build pipeline across macOS (arm64 + x64), Windows, and Linux.

latest

🧑‍💻 User Experience

🔥Computer Use
Give any agent hands on your desktop — no vision model required. Uses native Accessibility APIs to read and control any app directly, without screenshots. One-click enable for Claude Code, Codex, Gemini, OpenCode, and Pi. Agents can click UI elements, type, scroll, open apps, control Chrome via CDP, and fall back to OCR when needed. Built-in Computer Operator CoWorker + one-click permission setup. Runs natively on macOS (arm64 + x64), Windows x64, and Linux x64.

1. Multi-tab workspace with drag-reorder
Open unlimited agent tabs simultaneously, drag to reorder, wrap to multiple rows. Per-tab working directory with staged commit (Enter / Apply / Browse) and last-cwd persistence across restarts.

2. Unified agent history — browse & restore every conversation
Every session across all 9 agents is automatically captured and indexed. Browse the full history from the Logo menu, search by content, filter by agent, and restore any past conversation with one click — including the exact working directory and agent session ID, so the conversation resumes exactly where it left off.

3. Cross-agent usage analytics — tokens, cost, every session
Complete token and estimated cost tracking for every agent across every session. Dashboard shows total spend summary cards, a daily bar chart, a by-model breakdown table, and a by-agent breakdown table — all backed by a local SQLite store with 30+ model pricing. Filter by time period (today / 7d / 30d / 90d / custom) or by agent. Nothing leaves your machine.

4. 9 AI agent adapters
First-class support for Pi, Claude Code, Codex CLI, GitHub Copilot CLI, OpenCode, Gemini CLI, Qwen-Code, Crush, and Hermes — each with auto-detect, install, and per-agent accent color.

5. CoWorker system — role personas as system prompts
Reusable specialist personas injected as system prompts. Three source types: built-in (AGENT.md files), online (URL-cached from GitHub), custom (inline text / local file / URL). Includes 8 built-in engineer roles and 3 curated presets: Karpathy Code Guidelines, Devin-style Planner, Earnings Call Analyst.

6. Per-tab CoWorker selector
CoWorker dropdown in the CwdBar next to the working-directory input. Selecting a CoWorker immediately writes its system prompt into the running PTY — no tab restart required.

7. Cron job scheduler
Settings → Cron: schedule automated agent tasks with Interval / At Time / Custom cron modes, human-readable preview, enable/disable, clone, and delete. Dashboard shows next-run countdown, run count, and last-status per job.

8. Agent-native session resume
Conversation continuity across restarts: Claude Code (--resume), Codex (resume <id>), OpenCode (--session). Session ID is captured from agent output and restored on tab reopen.

9. Tab session persistence
All open tabs (id, title, agent, cwd) are saved on every change and fully restored on next launch, with the last-active tab focused first.

10. "Keep Awake" power management
One-click toggle in the Logo menu to block system sleep while long-running agent jobs are in progress — without affecting screen dimming.

11. In-app update checker
Polls GitHub Releases API 10 s after launch, then every 30 min. Green dot badge when a new version is available, with links to release notes and download.

⚙️ Configuration & Integration

12. 28-provider settings panel
CRUD UI for provider profiles covering DeepSeek, OpenAI, Anthropic, Google Gemini, xAI, Groq, Mistral, Ollama, LM Studio, OpenRouter, and 18 more — with OpenAI-compatible and Anthropic-compatible base URL modes.

13. Per-agent provider & model binding
Each agent tab can use a different provider and model. Model flags are projected to CLI arguments automatically at spawn time. "Use one provider for all agents" shared-config toggle also available.

14. Local inference auto-discovery
TCP pre-filter + HTTP fingerprint scan for Ollama, LM Studio, vLLM, llama.cpp, SGLang, LocalAI, and Jan. Discovered models appear as chips in the provider UI. Manual base-URL probe also available.

15. Reliable initialPrompt delivery
Task prompts reach the agent at spawn time even when the screen is locked. CLI positional arg for Codex / Claude Code / OpenCode / Gemini / Qwen-Code; bracketed-paste PTY write (ESC[200~) for Pi / Copilot / Crush / Hermes.

🔧 Technical Foundation

16. Real PTY terminal with xterm.js
Full bidirectional PTY via node-pty + xterm.js with fit, resize, and web-links addons. Correct COLUMNS/LINES env vars and RAF-timed spawn ensure accurate terminal dimensions from the first keystroke.

17. DeepSeek Anthropic gateway proxy
In-process HTTP server translating OpenAI Responses API → Anthropic Messages API — enabling streaming, extended thinking, tool use, and multi-turn conversations with DeepSeek models through Claude Code.

18. Resizable settings dialog
Drag handle in the bottom-right corner. Dialog size is lazy-initialized to fit the current viewport, so the tab row is never pushed out of the visible area.

19. Electron drag-region safety
All interactive elements (tab buttons, dropdowns, settings tabs, backdrop) are marked no-drag (-webkit-app-region: no-drag), preventing Electron's window-drag layer from swallowing clicks — especially when the TabBar wraps to multiple rows.

20. Windows & cross-platform support
PATH augmentation for nvm-windows, Volta, fnm, and other version managers. Native menu preserved on Windows. Consistent PTY column/row handling and build pipeline across macOS (arm64 + x64), Windows, and Linux.

🧑‍💻 User Experience

1. Multi-tab workspace with drag-reorder
Open unlimited agent tabs simultaneously, drag to reorder, wrap to multiple rows. Per-tab working directory with staged commit (Enter / Apply / Browse) and last-cwd persistence across restarts.

2. Unified agent history — browse & restore every conversation
Every session across all 9 agents is automatically captured and indexed. Browse the full history from the Logo menu, search by content, filter by agent, and restore any past conversation with one click — including the exact working directory and agent session ID, so the conversation resumes exactly where it left off.

3. Cross-agent usage analytics — tokens, cost, every session
Complete token and estimated cost tracking for every agent across every session. Dashboard shows total spend summary cards, a daily bar chart, a by-model breakdown table, and a by-agent breakdown table — all backed by a local SQLite store with 30+ model pricing. Filter by time period (today / 7d / 30d / 90d / custom) or by agent. Nothing leaves your machine.

4. 9 AI agent adapters
First-class support for Pi, Claude Code, Codex CLI, GitHub Copilot CLI, OpenCode, Gemini CLI, Qwen-Code, Crush, and Hermes — each with auto-detect, install, and per-agent accent color.

5. CoWorker system — role personas as system prompts
Reusable specialist personas injected as system prompts. Three source types: built-in (AGENT.md files), online (URL-cached from GitHub), custom (inline text / local file / URL). Includes 8 built-in engineer roles and 3 curated presets: Karpathy Code Guidelines, Devin-style Planner, Earnings Call Analyst.

6. Per-tab CoWorker selector
CoWorker dropdown in the CwdBar next to the working-directory input. Selecting a CoWorker immediately writes its system prompt into the running PTY — no tab restart required.

7. Cron job scheduler
Settings → Cron: schedule automated agent tasks with Interval / At Time / Custom cron modes, human-readable preview, enable/disable, clone, and delete. Dashboard shows next-run countdown, run count, and last-status per job.

8. Agent-native session resume
Conversation continuity across restarts: Claude Code (--resume), Codex (resume <id>), OpenCode (--session). Session ID is captured from agent output and restored on tab reopen.

9. Tab session persistence
All open tabs (id, title, agent, cwd) are saved on every change and fully restored on next launch, with the last-active tab focused first.

10. "Keep Awake" power management
One-click toggle in the Logo menu to block system sleep while long-running agent jobs are in progress — without affecting screen dimming.

11. In-app update checker
Polls GitHub Releases API 10 s after launch, then every 30 min. Green dot badge when a new version is available, with links to release notes and download.

⚙️ Configuration & Integration

12. 28-provider settings panel
CRUD UI for provider profiles covering DeepSeek, OpenAI, Anthropic, Google Gemini, xAI, Groq, Mistral, Ollama, LM Studio, OpenRouter, and 18 more — with OpenAI-compatible and Anthropic-compatible base URL modes.

13. Per-agent provider & model binding
Each agent tab can use a different provider and model. Model flags are projected to CLI arguments automatically at spawn time. "Use one provider for all agents" shared-config toggle also available.

14. Local inference auto-discovery
TCP pre-filter + HTTP fingerprint scan for Ollama, LM Studio, vLLM, llama.cpp, SGLang, LocalAI, and Jan. Discovered models appear as chips in the provider UI. Manual base-URL probe also available.

15. Reliable initialPrompt delivery
Task prompts reach the agent at spawn time even when the screen is locked. CLI positional arg for Codex / Claude Code / OpenCode / Gemini / Qwen-Code; bracketed-paste PTY write (ESC[200~) for Pi / Copilot / Crush / Hermes.

🔧 Technical Foundation

16. Real PTY terminal with xterm.js
Full bidirectional PTY via node-pty + xterm.js with fit, resize, and web-links addons. Correct COLUMNS/LINES env vars and RAF-timed spawn ensure accurate terminal dimensions from the first keystroke.

17. DeepSeek Anthropic gateway proxy
In-process HTTP server translating OpenAI Responses API → Anthropic Messages API — enabling streaming, extended thinking, tool use, and multi-turn conversations with DeepSeek models through Claude Code.

18. Resizable settings dialog
Drag handle in the bottom-right corner. Dialog size is lazy-initialized to fit the current viewport, so the tab row is never pushed out of the visible area.

19. Electron drag-region safety
All interactive elements (tab buttons, dropdowns, settings tabs, backdrop) are marked no-drag (-webkit-app-region: no-drag), preventing Electron's window-drag layer from swallowing clicks — especially when the TabBar wraps to multiple rows.

20. Windows & cross-platform support
PATH augmentation for nvm-windows, Volta, fnm, and other version managers. Native menu preserved on Windows. Consistent PTY column/row handling and build pipeline across macOS (arm64 + x64), Windows, and Linux.

🧑‍💻 User Experience

1. Multi-tab workspace with drag-reorder
Open unlimited agent tabs simultaneously, drag to reorder, wrap to multiple rows. Per-tab working directory with staged commit (Enter / Apply / Browse) and last-cwd persistence across restarts.

2. Unified agent history — browse & restore every conversation
Every session across all 9 agents is automatically captured and indexed. Browse the full history from the Logo menu, search by content, filter by agent, and restore any past conversation with one click — including the exact working directory and agent session ID, so the conversation resumes exactly where it left off.

3. Cross-agent usage analytics — tokens, cost, every session
Complete token and estimated cost tracking for every agent across every session. Dashboard shows total spend summary cards, a daily bar chart, a by-model breakdown table, and a by-agent breakdown table — all backed by a local SQLite store with 30+ model pricing. Filter by time period (today / 7d / 30d / 90d / custom) or by agent. Nothing leaves your machine.

4. 9 AI agent adapters
First-class support for Pi, Claude Code, Codex CLI, GitHub Copilot CLI, OpenCode, Gemini CLI, Qwen-Code, Crush, and Hermes — each with auto-detect, install, and per-agent accent color.

5. CoWorker system — role personas as system prompts
Reusable specialist personas injected as system prompts. Three source types: built-in (AGENT.md files), online (URL-cached from GitHub), custom (inline text / local file / URL). Includes 8 built-in engineer roles and 3 curated presets: Karpathy Code Guidelines, Devin-style Planner, Earnings Call Analyst.

6. Per-tab CoWorker selector
CoWorker dropdown in the CwdBar next to the working-directory input. Selecting a CoWorker immediately writes its system prompt into the running PTY — no tab restart required.

7. Cron job scheduler
Settings → Cron: schedule automated agent tasks with Interval / At Time / Custom cron modes, human-readable preview, enable/disable, clone, and delete. Dashboard shows next-run countdown, run count, and last-status per job.

8. Agent-native session resume
Conversation continuity across restarts: Claude Code (--resume), Codex (resume <id>), OpenCode (--session). Session ID is captured from agent output and restored on tab reopen.

9. Tab session persistence
All open tabs (id, title, agent, cwd) are saved on every change and fully restored on next launch, with the last-active tab focused first.

10. "Keep Awake" power management
One-click toggle in the Logo menu to block system sleep while long-running agent jobs are in progress — without affecting screen dimming.

11. In-app update checker
Polls GitHub Releases API 10 s after launch, then every 30 min. Green dot badge when a new version is available, with links to release notes and download.

⚙️ Configuration & Integration

12. 28-provider settings panel
CRUD UI for provider profiles covering DeepSeek, OpenAI, Anthropic, Google Gemini, xAI, Groq, Mistral, Ollama, LM Studio, OpenRouter, and 18 more — with OpenAI-compatible and Anthropic-compatible base URL modes.

13. Per-agent provider & model binding
Each agent tab can use a different provider and model. Model flags are projected to CLI arguments automatically at spawn time. "Use one provider for all agents" shared-config toggle also available.

14. Local inference auto-discovery
TCP pre-filter + HTTP fingerprint scan for Ollama, LM Studio, vLLM, llama.cpp, SGLang, LocalAI, and Jan. Discovered models appear as chips in the provider UI. Manual base-URL probe also available.

15. Reliable initialPrompt delivery
Task prompts reach the agent at spawn time even when the screen is locked. CLI positional arg for Codex / Claude Code / OpenCode / Gemini / Qwen-Code; bracketed-paste PTY write (ESC[200~) for Pi / Copilot / Crush / Hermes.

🔧 Technical Foundation

16. Real PTY terminal with xterm.js
Full bidirectional PTY via node-pty + xterm.js with fit, resize, and web-links addons. Correct COLUMNS/LINES env vars and RAF-timed spawn ensure accurate terminal dimensions from the first keystroke.

17. DeepSeek Anthropic gateway proxy
In-process HTTP server translating OpenAI Responses API → Anthropic Messages API — enabling streaming, extended thinking, tool use, and multi-turn conversations with DeepSeek models through Claude Code.

18. Resizable settings dialog
Drag handle in the bottom-right corner. Dialog size is lazy-initialized to fit the current viewport, so the tab row is never pushed out of the visible area.

19. Electron drag-region safety
All interactive elements (tab buttons, dropdowns, settings tabs, backdrop) are marked no-drag (-webkit-app-region: no-drag), preventing Electron's window-drag layer from swallowing clicks — especially when the TabBar wraps to multiple rows.

20. Windows & cross-platform support
PATH augmentation for nvm-windows, Volta, fnm, and other version managers. Native menu preserved on Windows. Consistent PTY column/row handling and build pipeline across macOS (arm64 + x64), Windows, and Linux.

🧑‍💻 User Experience

1. Multi-tab workspace with drag-reorder
Open unlimited agent tabs simultaneously, drag to reorder, wrap to multiple rows. Per-tab working directory with staged commit (Enter / Apply / Browse) and last-cwd persistence across restarts.

2. Unified agent history — browse & restore every conversation
Every session across all 9 agents is automatically captured and indexed. Browse the full history from the Logo menu, search by content, filter by agent, and restore any past conversation with one click — including the exact working directory and agent session ID, so the conversation resumes exactly where it left off.

3. Cross-agent usage analytics — tokens, cost, every session
Complete token and estimated cost tracking for every agent across every session. Dashboard shows total spend summary cards, a daily bar chart, a by-model breakdown table, and a by-agent breakdown table — all backed by a local SQLite store with 30+ model pricing. Filter by time period (today / 7d / 30d / 90d / custom) or by agent. Nothing leaves your machine.

4. 9 AI agent adapters
First-class support for Pi, Claude Code, Codex CLI, GitHub Copilot CLI, OpenCode, Gemini CLI, Qwen-Code, Crush, and Hermes — each with auto-detect, install, and per-agent accent color.

5. CoWorker system — role personas as system prompts
Reusable specialist personas injected as system prompts. Three source types: built-in (AGENT.md files), online (URL-cached from GitHub), custom (inline text / local file / URL). Includes 8 built-in engineer roles and 3 curated presets: Karpathy Code Guidelines, Devin-style Planner, Earnings Call Analyst.

6. Per-tab CoWorker selector
CoWorker dropdown in the CwdBar next to the working-directory input. Selecting a CoWorker immediately writes its system prompt into the running PTY — no tab restart required.

7. Cron job scheduler
Settings → Cron: schedule automated agent tasks with Interval / At Time / Custom cron modes, human-readable preview, enable/disable, clone, and delete. Dashboard shows next-run countdown, run count, and last-status per job.

8. Agent-native session resume
Conversation continuity across restarts: Claude Code (--resume), Codex (resume <id>), OpenCode (--session). Session ID is captured from agent output and restored on tab reopen.

9. Tab session persistence
All open tabs (id, title, agent, cwd) are saved on every change and fully restored on next launch, with the last-active tab focused first.

10. "Keep Awake" power management
One-click toggle in the Logo menu to block system sleep while long-running agent jobs are in progress — without affecting screen dimming.

11. In-app update checker
Polls GitHub Releases API 10 s after launch, then every 30 min. Green dot badge when a new version is available, with links to release notes and download.

⚙️ Configuration & Integration

12. 28-provider settings panel
CRUD UI for provider profiles covering DeepSeek, OpenAI, Anthropic, Google Gemini, xAI, Groq, Mistral, Ollama, LM Studio, OpenRouter, and 18 more — with OpenAI-compatible and Anthropic-compatible base URL modes.

13. Per-agent provider & model binding
Each agent tab can use a different provider and model. Model flags are projected to CLI arguments automatically at spawn time. "Use one provider for all agents" shared-config toggle also available.

14. Local inference auto-discovery
TCP pre-filter + HTTP fingerprint scan for Ollama, LM Studio, vLLM, llama.cpp, SGLang, LocalAI, and Jan. Discovered models appear as chips in the provider UI. Manual base-URL probe also available.

15. Reliable initialPrompt delivery
Task prompts reach the agent at spawn time even when the screen is locked. CLI positional arg for Codex / Claude Code / OpenCode / Gemini / Qwen-Code; bracketed-paste PTY write (ESC[200~) for Pi / Copilot / Crush / Hermes.

🔧 Technical Foundation

16. Real PTY terminal with xterm.js
Full bidirectional PTY via node-pty + xterm.js with fit, resize, and web-links addons. Correct COLUMNS/LINES env vars and RAF-timed spawn ensure accurate terminal dimensions from the first keystroke.

17. DeepSeek Anthropic gateway proxy
In-process HTTP server translating OpenAI Responses API → Anthropic Messages API — enabling streaming, extended thinking, tool use, and multi-turn conversations with DeepSeek models through Claude Code.

18. Resizable settings dialog
Drag handle in the bottom-right corner. Dialog size is lazy-initialized to fit the current viewport, so the tab row is never pushed out of the visible area.

19. Electron drag-region safety
All interactive elements (tab buttons, dropdowns, settings tabs, backdrop) are marked no-drag (-webkit-app-region: no-drag), preventing Electron's window-drag layer from swallowing clicks — especially when the TabBar wraps to multiple rows.

20. Windows & cross-platform support
PATH augmentation for nvm-windows, Volta, fnm, and other version managers. Native menu preserved on Windows. Consistent PTY column/row handling and build pipeline across macOS (arm64 + x64), Windows, and Linux.

Full Changelog: v0.7.31...v0.7.32

🧑‍💻 User Experience

1. Multi-tab workspace with drag-reorder
Open unlimited agent tabs simultaneously, drag to reorder, wrap to multiple rows. Per-tab working directory with staged commit (Enter / Apply / Browse) and last-cwd persistence across restarts.

2. Unified agent history — browse & restore every conversation
Every session across all 9 agents is automatically captured and indexed. Browse the full history from the Logo menu, search by content, filter by agent, and restore any past conversation with one click — including the exact working directory and agent session ID, so the conversation resumes exactly where it left off.

3. Cross-agent usage analytics — tokens, cost, every session
Complete token and estimated cost tracking for every agent across every session. Dashboard shows total spend summary cards, a daily bar chart, a by-model breakdown table, and a by-agent breakdown table — all backed by a local SQLite store with 30+ model pricing. Filter by time period (today / 7d / 30d / 90d / custom) or by agent. Nothing leaves your machine.

4. 9 AI agent adapters
First-class support for Pi, Claude Code, Codex CLI, GitHub Copilot CLI, OpenCode, Gemini CLI, Qwen-Code, Crush, and Hermes — each with auto-detect, install, and per-agent accent color.

5. CoWorker system — role personas as system prompts
Reusable specialist personas injected as system prompts. Three source types: built-in (AGENT.md files), online (URL-cached from GitHub), custom (inline text / local file / URL). Includes 8 built-in engineer roles and 3 curated presets: Karpathy Code Guidelines, Devin-style Planner, Earnings Call Analyst.

6. Per-tab CoWorker selector
CoWorker dropdown in the CwdBar next to the working-directory input. Selecting a CoWorker immediately writes its system prompt into the running PTY — no tab restart required.

7. Cron job scheduler
Settings → Cron: schedule automated agent tasks with Interval / At Time / Custom cron modes, human-readable preview, enable/disable, clone, and delete. Dashboard shows next-run countdown, run count, and last-status per job.

8. Agent-native session resume
Conversation continuity across restarts: Claude Code (--resume), Codex (resume <id>), OpenCode (--session). Session ID is captured from agent output and restored on tab reopen.

9. Tab session persistence
All open tabs (id, title, agent, cwd) are saved on every change and fully restored on next launch, with the last-active tab focused first.

10. "Keep Awake" power management
One-click toggle in the Logo menu to block system sleep while long-running agent jobs are in progress — without affecting screen dimming.

11. In-app update checker
Polls GitHub Releases API 10 s after launch, then every 30 min. Green dot badge when a new version is available, with links to release notes and download.

⚙️ Configuration & Integration

12. 28-provider settings panel
CRUD UI for provider profiles covering DeepSeek, OpenAI, Anthropic, Google Gemini, xAI, Groq, Mistral, Ollama, LM Studio, OpenRouter, and 18 more — with OpenAI-compatible and Anthropic-compatible base URL modes.

13. Per-agent provider & model binding
Each agent tab can use a different provider and model. Model flags are projected to CLI arguments automatically at spawn time. "Use one provider for all agents" shared-config toggle also available.

14. Local inference auto-discovery
TCP pre-filter + HTTP fingerprint scan for Ollama, LM Studio, vLLM, llama.cpp, SGLang, LocalAI, and Jan. Discovered models appear as chips in the provider UI. Manual base-URL probe also available.

15. Reliable initialPrompt delivery
Task prompts reach the agent at spawn time even when the screen is locked. CLI positional arg for Codex / Claude Code / OpenCode / Gemini / Qwen-Code; bracketed-paste PTY write (ESC[200~) for Pi / Copilot / Crush / Hermes.

🔧 Technical Foundation

16. Real PTY terminal with xterm.js
Full bidirectional PTY via node-pty + xterm.js with fit, resize, and web-links addons. Correct COLUMNS/LINES env vars and RAF-timed spawn ensure accurate terminal dimensions from the first keystroke.

17. DeepSeek Anthropic gateway proxy
In-process HTTP server translating OpenAI Responses API → Anthropic Messages API — enabling streaming, extended thinking, tool use, and multi-turn conversations with DeepSeek models through Claude Code.

18. Resizable settings dialog
Drag handle in the bottom-right corner. Dialog size is lazy-initialized to fit the current viewport, so the tab row is never pushed out of the visible area.

19. Electron drag-region safety
All interactive elements (tab buttons, dropdowns, settings tabs, backdrop) are marked no-drag (-webkit-app-region: no-drag), preventing Electron's window-drag layer from swallowing clicks — especially when the TabBar wraps to multiple rows.

20. Windows & cross-platform support
PATH augmentation for nvm-windows, Volta, fnm, and other version managers. Native menu preserved on Windows. Consistent PTY column/row handling and build pipeline across macOS (arm64 + x64), Windows, and Linux.