CLI

lacause edited this page Mar 29, 2026 · 2 revisions

CLI Reference

OCC includes a command-line interface for development, validation, and execution.

Installation

The CLI is included when you install OCC:

cd OCC/mcp-server
npm install && npm run build
# Run via npm
npm run occ -- <command>
# Or directly
node dist/bin/occ.js <command>

Commands

occ list

List all available chains and pipelines.

$ occ list
Chains:
 deep-researcher (6 steps) — Multi-perspective research engine
 code-review (8 steps) — Intelligent code review pipeline
 content-engine (6 steps) — End-to-end content creation
 competitive-intel (4 steps) — Competitive intelligence analysis
 full-stack-scaffold (5 steps) — AI-powered project scaffolder
 market-monitor (4 steps) — Real-time market monitoring
Pipelines:
 research-to-content (2 chains) — Research then write article
 product-intelligence (2 chains) — Competitive analysis + market alerts

occ validate [path]

Lint and validate all chains in a directory. Catches:

Undefined variable references
Unreachable steps
Missing depends_on targets
Invalid router routes
Missing pre-tool fields
Unused output variables
Circular dependencies
Duplicate step IDs

$ occ validate ./chains
 ✓ code-review (8 steps)
 ✓ competitive-intel (4 steps)
 ⚠ content-engine (6 steps)
 info social_media: output_var "social_posts" is never referenced
 ✓ deep-researcher (6 steps)
 ✓ market-monitor (4 steps)
6 chains validated: 0 errors, 0 warnings

Exit code is 1 if any errors are found — perfect for CI:

# .github/workflows/ci.yml
- name: Validate chains
 run: cd mcp-server && node dist/bin/occ.js validate ../chains

occ dry-run <chain> [--input k=v]

Preview the execution plan without making any LLM calls. Shows:

Execution waves (parallel groups)
Prompt previews with resolved variables
Model assignments per step
Dependency graph
Cost estimates

$ occ dry-run deep-researcher --input topic="quantum computing"
Execution Plan: deep-researcher
 Wave 1 (3 parallel)
 search_mainstream [claude-sonnet-4-6]
 search_contrarian [claude-sonnet-4-6]
 search_academic [claude-sonnet-4-6]
 Wave 2
 evaluate_sources [claude-sonnet-4-6] ← search_mainstream, ...
 Wave 3
 merge_perspectives [claude-sonnet-4-6]
 Wave 4
 synthesize [claude-sonnet-4-6]
Estimated Cost:
 Steps: 6 (4 waves)
 Models: claude-sonnet-4-6: 6 steps (~0ドル.126-0ドル.360)
 Total: ~0ドル.126-0ドル.360

occ run <chain> [--input k=v ...]

Execute a chain and stream real-time logs.

$ occ run deep-researcher --input topic="quantum computing"
Executing: deep-researcher
Inputs: {"topic":"quantum computing"}
Execution: 1a2b3c4d5e6f
[START] Chain: deep-researcher
 [STEP] Mainstream perspective ...
 [STEP] Contrarian perspective ...
 [STEP] Academic perspective ...
 [DONE] search_mainstream (12.3s, 450+1800 tokens)
 [DONE] search_contrarian (14.1s, 380+1600 tokens)
 [DONE] search_academic (11.8s, 420+1900 tokens)
 [STEP] Source reliability check ...
 [DONE] evaluate_sources (5.2s)
 [STEP] Merge all research ...
 [DONE] merge_perspectives (8.7s)
 [STEP] Final report ...
 [DONE] synthesize (15.4s)
--- Result ---
# Research Report: Quantum Computing ...

occ status <executionId>

Check the status of an execution.

$ occ status 1a2b3c4d
Execution: 1a2b3c4d
Chain: deep-researcher
Status: done
Duration: 47.2s
Steps:
 done search_mainstream (12.3s) [450+1800 tokens]
 done search_contrarian (14.1s) [380+1600 tokens]
 done search_academic (11.8s) [420+1900 tokens]
 done evaluate_sources (5.2s)
 done merge_perspectives (8.7s)
 done synthesize (15.4s)

occ logs <executionId>

Stream real-time SSE events for a running or completed execution.

$ occ logs 1a2b3c4d

occ cancel <executionId>

Cancel a running execution. Kills all active processes.

$ occ cancel 1a2b3c4d
Cancelled: 1a2b3c4d

occ queue

Show queue statistics and recent jobs.

$ occ queue
Queue Stats:
 Queued: 3
 Running: 2
 Done: 47
 Errors: 1
 Workers: 2/5
 Avg wait: 4s
Recent Jobs:
 done deep-researcher (p5) → exec_abc123
 running code-review (p10)
 queued content-engine (p3)

occ timeline <executionId>

Time-travel: view the full step checkpoint history.

$ occ timeline 1a2b3c4d
Timeline for 1a2b3c4d:
 done search_mainstream (12.3s) [450+1800 tok] 2026年03月29日 14:00:12
 done search_contrarian (14.1s) [380+1600 tok] 2026年03月29日 14:00:14
 done search_academic (11.8s) [420+1900 tok] 2026年03月29日 14:00:12
 done evaluate_sources (5.2s) 2026年03月29日 14:00:26
 done merge_perspectives (8.7s) 2026年03月29日 14:00:31
 done synthesize (15.4s) 2026年03月29日 14:00:40

occ stats <chainName>

Execution statistics for a chain (from SQLite).

$ occ stats deep-researcher
Stats for deep-researcher:
 Total runs: 12
 Success rate: 91.7%
 Avg duration: 42.3s
 Total tokens: 14500 input, 38200 output
 Est. total cost: ~0ドル.617 (Sonnet rates)

occ approve / reject

Approve or reject a gate step.

$ occ approve 1a2b3c4d gate_step
Approved: 1a2b3c4d / gate_step
$ occ reject 1a2b3c4d gate_step
Rejected: 1a2b3c4d / gate_step

occ run-pipeline

Execute a pipeline (multi-chain workflow).

$ occ run-pipeline research-to-content --input topic="AI"
Executing pipeline: research-to-content
Inputs: {"topic":"AI"}
Execution: pip_abc123

occ health

Check server health, queue status, and MCP servers.

$ occ health
OCC server is running
 Version: 2.0.0
 Running executions: 2
 Queue: 3 queued, 2 running, 47 done
 MCP servers: github, brave-search
 URL: http://localhost:4242

Global Flags

Flag	Description
`--json`	Output raw JSON (for scripting, piping to `jq`, CI/CD)
`--priority N`, `-p N`	Execution priority 1-10 (default: 5, higher = first)
`--input k=v`, `-i k=v`	Chain/pipeline input (repeatable)

# Machine-readable output
occ stats deep-researcher --json | jq '.successRate'
# High-priority execution
occ run urgent-chain --input data="critical" --priority 10

Environment Variables

Variable	Default	Description
`OCC_URL`	`http://localhost:4242`	Server URL for run/status/logs/health
`CHAINS_DIR`	`../chains`	Chains directory for validate/dry-run
`REST_PORT`	`4242`	Server port (used to build OCC_URL)

Cost Estimation

Dry-run includes cost estimates based on average token usage per step:

Model	Input cost/M	Output cost/M	Est. per step
claude-haiku-4-5	0ドル.25	1ドル.25	0ドル.002-0ドル.005
claude-sonnet-4-6	3ドル.00	15ドル.00	0ドル.021-0ドル.060
claude-opus-4-6	15ドル.00	75ドル.00	0ドル.105-0ドル.300

Estimates assume 2K-5K input tokens and 1K-3K output tokens per step.

CLI

CLI Reference

Installation

Commands

occ list

occ validate [path]

occ dry-run <chain> [--input k=v]

occ run <chain> [--input k=v ...]

occ status <executionId>

occ logs <executionId>

occ cancel <executionId>

occ queue

occ timeline <executionId>

occ stats <chainName>

occ approve / reject

occ run-pipeline

occ health

Global Flags

Environment Variables

Cost Estimation

See Also

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Clone this wiki locally