What's Changed

• chore: docs updates for latest v0.2.X release by @Aakashbhardwaj27 in #8
• feat: add tolling by @Aakashbhardwaj27 in #9
• fix: fixed the codecov setup by @Aakashbhardwaj27 in #10
• Dev by @Aakashbhardwaj27 in #11
• fix: updated codecov action setup by @Aakashbhardwaj27 in #19
• fix: updated codecov action setup by @Aakashbhardwaj27 in #20
• chore: updated dependency review for dev also by @Aakashbhardwaj27 in #21
• chore: v0.2.3 release by @Aakashbhardwaj27 in #22

Full Changelog: v0.2.2...v0.2.3

What's Changed

• chore: docs updates for latest v0.2.X release by @Aakashbhardwaj27 in #8
• feat: add tolling by @Aakashbhardwaj27 in #9
• fix: fixed the codecov setup by @Aakashbhardwaj27 in #10
• Dev by @Aakashbhardwaj27 in #11
• fix: updated codecov action setup by @Aakashbhardwaj27 in #19
• fix: updated codecov action setup by @Aakashbhardwaj27 in #20
• chore: updated dependency review for dev also by @Aakashbhardwaj27 in #21
• chore: v0.2.3 release by @Aakashbhardwaj27 in #22

Full Changelog: v0.2.2...v0.2.3

Contributors

• @sky-in-code

What's Changed

• fix: version bumpup by @Aakashbhardwaj27 in #7

Full Changelog: v0.2.1...v0.2.2

What's Changed

• fix: version bumpup by @Aakashbhardwaj27 in #7

Full Changelog: v0.2.1...v0.2.2

Contributors

• @sky-in-code

What's Changed

• feat: updated content for v0.2.0 by @Aakashbhardwaj27 in #5
• feat: added husky by @Aakashbhardwaj27 in #6

Full Changelog: v0.1.1...v0.2.1

Contributors

• @sky-in-code

What's Changed

• feat: updated content for v0.2.0 by @Aakashbhardwaj27 in #5

Full Changelog: v0.1.1...v0.2.0-alpha.1

What's Changed

• feat: updated content for v0.2.0 by @Aakashbhardwaj27 in #5

Full Changelog: v0.1.1...v0.2.0-alpha.1

Contributors

• @sky-in-code

Full Changelog: v0.1.0-alpha.3...v0.1.1

Full Changelog: v0.1.0-alpha.3...v0.1.1

v0.1.0-alpha.2 — Genesis

The first public release of AgentVM — the runtime your AI agents deserve.

Install

npm install @llmhut/agentvm@alpha

What's in this release

AgentVM is a framework-agnostic runtime for autonomous AI agents. It provides the infrastructure layer that sits beneath agent frameworks like LangChain, CrewAI, and AutoGen — handling process lifecycle, memory, tools, messaging, and scheduling so framework developers can focus on reasoning and workflow design.

Core Runtime

• Kernel — central orchestrator with agent registration, process lifecycle, and event dispatch
• Agent — typed definitions with handler functions, tool declarations, and memory config
• Process — full state machine (created → running → paused → terminated / crashed)
• kernel.execute() — builds an ExecutionContext with memory, tools, and messaging, then calls the agent's handler

Memory Bus

• Per-process isolated working memory
• Cross-agent shared memory
• Automatic cleanup on terminate (or persist with memory.persistent: true)

Tool Router

• Register tools with typed schemas and permission levels
• Rate limiting per tool per agent
• Agents can only call tools they declare in their config

Message Broker

• Pub/sub channels for inter-agent communication
• Direct messaging between specific processes
• Channel history with configurable limits

Scheduler

• Sequential, parallel, race, and conditional execution strategies
• Dependency resolution with cycle detection
• Retry with fixed or exponential backoff

CLI

• agentvm init — scaffold a new project in seconds
• agentvm start — interactive console with spawn, exec, ps, kill, pause, resume, logs, stats
• Colored terminal output with state badges and formatted tables

Developer Experience

• Full TypeScript with strict mode
• 135 unit tests at 90%+ coverage
• 3 example projects
• Architecture docs, Getting Started guide, and RFCs

Quick Start

import { Kernel, Agent } from '@llmhut/agentvm';
const kernel = new Kernel();
const agent = new Agent({
 name: 'greeter',
 handler: async (ctx) => {
 const count = ((await ctx.memory.get('count')) as number ?? 0) + 1;
 await ctx.memory.set('count', count);
 return `Hello, ${ctx.input}! (greeting #${count})`;
 },
});
kernel.register(agent);
const proc = await kernel.spawn('greeter');
const result = await kernel.execute(proc.id, { task: 'World' });
console.log(result.output); // "Hello, World! (greeting #1)"

What's next

Phase 2 (v0.2.0) — pluggable memory backends (SQLite, Redis), agent contracts with runtime validation, and YAML configuration system. See ROADMAP.md for the full plan.

Links: GitHub · npm · Discord

Full Changelog: v0.1.0-alpha.2...v0.1.0-alpha.3

v0.1.0-alpha.1 — Genesis

The first public release of AgentVM — the runtime your AI agents deserve.

What's in this release

AgentVM is a framework-agnostic runtime for autonomous AI agents.
It provides the infrastructure layer that sits beneath agent
frameworks like LangChain, CrewAI, and AutoGen.

Core Runtime

• Kernel — central orchestrator with agent registration, process
  lifecycle, and event dispatch
• Agent — typed definitions with handler functions, tool
  declarations, and memory config
• Process — full state machine
  (created → running → paused → terminated/crashed)
• kernel.execute() — builds ExecutionContext with memory, tools,
  messaging, then calls the agent handler

Memory Bus

• Per-process isolated working memory
• Cross-agent shared memory
• Automatic cleanup on terminate (or persist with memory.persistent: true)

Tool Router

• Register tools with typed schemas and permission levels
• Rate limiting per tool per agent
• Agents can only call tools they declare in their config

Message Broker

• Pub/sub channels for inter-agent communication
• Direct messaging between specific processes
• Channel history with configurable limits

Scheduler

• Sequential, parallel, race, and conditional execution strategies
• Dependency resolution with cycle detection
• Retry with fixed or exponential backoff

CLI

• agentvm init — scaffold a new project in seconds
• agentvm start — interactive console with spawn, exec, ps,
  kill, pause, resume, logs, stats

Developer Experience

• Full TypeScript with strict mode
• 135 unit tests at 90%+ coverage
• 3 example projects
• Architecture docs and RFCs

Install

npm install agentvm

Quick Start

import { Kernel, Agent } from 'agentvm';
const kernel = new Kernel();
const agent = new Agent({
 name: 'greeter',
 handler: async (ctx) => {
 await ctx.memory.set('count', 
 ((await ctx.memory.get('count')) as number ?? 0) + 1
 );
 return `Hello, ${ctx.input}!`;
 },
});
kernel.register(agent);
const proc = await kernel.spawn('greeter');
const result = await kernel.execute(proc.id, { task: 'World' });
console.log(result.output); // "Hello, World!"

What's next

Phase 2 (v0.2.0) — pluggable memory backends (SQLite, Redis),
agent contracts with typed validation, YAML config system, and
framework adapter plugins for LangChain and CrewAI.

See ROADMAP.md
for the full plan.

Built by @llmhut.
Star the repo if this resonates — let's build the foundation
of agentic AI together.