The cognitive layer for multi-agent systems.
Give your agents a shared brain.
Website · Docs · Quick Start · Roadmap · Contributing
Your agents talk to LLMs. Why don't they talk to each other?
You have orchestration. You have tools. What you don't have is a place where agents validate each other's findings, build on shared context, and learn from what actually happened in production.
AMFS connects your agents across frameworks, sessions, and machines. They share knowledge, coordinate, and learn as a team.
Discover — Before acting, an agent queries AMFS for what the fleet already knows. Results come back ranked by confidence with full provenance — which agent wrote them, when, and how trustworthy they are.
Handoff — A finding written by one agent is readable by every other agent within milliseconds. The Cursor agent on your laptop and the LangGraph pipeline in production share the same memory.
Learn — When a deploy succeeds or an incident fires, AMFS records what the agent read, what it chose, and what happened. Confidence shifts based on real outcomes. Your training data writes itself.
pip install amfsfrom amfs import AgentMemory, OutcomeType
mem = AgentMemory(agent_id="research-agent")
# Agent discovers a pattern and commits it to memory
mem.write("checkout-service", "retry-pattern",
{"max_retries": 3, "strategy": "exponential-backoff"},
confidence=0.85)
# Another agent reads it — the read is tracked automatically
entry = mem.read("checkout-service", "retry-pattern")
# Get a compiled briefing: everything the fleet knows about this entity
briefing = mem.briefing("checkout-service")
# When the deploy succeeds, confidence on related entries adjusts
mem.commit_outcome("DEP-287", OutcomeType.SUCCESS)
# Explain exactly which memories and contexts drove a decision
trace = mem.explain("DEP-287")
Drop AMFS into whatever you're already using. No rewrites. Your agents start sharing what they know.
MCP — Cursor, Claude Desktop, Claude Code
One command to give any MCP-compatible client persistent agent memory:
curl -sSL https://raw.githubusercontent.com/raia-live/amfs/main/install-mcp.sh | bashOr add manually to your MCP config:
{
"mcpServers": {
"amfs": {
"command": "python",
"args": ["-m", "amfs_mcp_server"],
"env": { "AMFS_API_KEY": "sk-your-key-here" }
}
}
}Python SDK
pip install amfsfrom amfs import AgentMemory
mem = AgentMemory(agent_id="my-agent")
mem.write("my-service", "finding", "mutex pattern applied", confidence=0.91)
briefing = mem.briefing("my-service")TypeScript SDK
npm install @senselab-ai/amfsimport { HttpAdapter } from '@senselab-ai/amfs';
const adapter = new HttpAdapter({ agentId: 'my-agent', apiKey: 'sk-...' });
await adapter.writeAsync('my-service', 'finding', 'mutex pattern applied', { confidence: 0.91 });
const briefing = await adapter.briefingAsync('my-service');Framework integrations — Strands, CrewAI, LangGraph, LangChain, AutoGen
pip install amfs-strands # AWS Strands Agents
pip install amfs-crewai # CrewAI
pip install amfs-langgraph # LangGraph
pip install amfs-langchain # LangChain
pip install amfs-autogen # AutoGen| Feature | What it does | |
|---|---|---|
| 🧠 | Confidence & Outcomes | Entries carry trust scores that shift when deploys succeed or incidents fire. |
| 🔍 | Briefings | One call returns everything the fleet knows about an entity — compiled, ranked, with provenance. |
| 📊 | Hybrid Search | Full-text + semantic + recency + confidence in a single ranked result set. |
| 🌳 | Branching & PRs | Create branches, diff changes, open pull requests, merge or discard — just like Git. |
| ⏪ | Rollback & Tags | Named snapshots. Restore to any tag or event. |
| 🔗 | Knowledge Graph | Relationships auto-materialize from normal operations. Query connections between entities. |
| 📖 | Causal Explainability | explain() shows exactly which memories and contexts drove a decision. |
| 🕐 | Git-like Timeline | Every write, outcome, and cross-agent read is logged. Full audit trail. |
| 🔒 | Access Control | Grant read or read/write per branch, user, team, or API key. Multi-tenant isolation. |
| 📦 | Training Signal Output | Export SFT/DPO datasets from real decision traces — not synthetic data. |
| 🔌 | MCP Server | First-class support for Cursor, Claude Desktop, Claude Code, and any MCP client. |
| 🧩 | Connectors | PagerDuty, GitHub, Slack, Jira — or build your own. |
AMFS is a monorepo with a layered architecture. Pick what you need:
| Package | Install | Description |
|---|---|---|
| Python SDK | pip install amfs |
Core client library — AgentMemory class |
| TypeScript SDK | npm install @senselab-ai/amfs |
TypeScript client with full async API |
| MCP Server | pip install amfs-mcp-server |
Model Context Protocol server for Cursor / Claude |
| HTTP Server | pip install amfs-http-server |
REST API server |
| Core Engine | pip install amfs-core |
Storage engine, versioning, timeline |
| CLI | pip install amfs-cli |
Command-line tools |
Storage adapters — plug in your backend:
| Adapter | Install |
|---|---|
| Filesystem (default) | pip install amfs-adapter-filesystem |
| PostgreSQL | pip install amfs-adapter-postgres |
| S3 | pip install amfs-adapter-s3 |
| HTTP (remote) | pip install amfs-adapter-http |
Integrations:
pip install amfs-strands # AWS Strands Agents
pip install amfs-crewai # CrewAI
pip install amfs-langgraph # LangGraph
pip install amfs-langchain # LangChain
pip install amfs-autogen # AutoGenOr run with Docker:
docker run -p 8080:8080 -v amfs-data:/data ghcr.io/raia-live/amfsAMFS is open source under Apache 2.0. The OSS edition gives you the full memory engine — versioned writes, confidence scoring, outcome feedback, causal traces, knowledge graph, hybrid search, git-like timeline, SDKs, adapters, HTTP API, MCP server, and CLI.
AMFS Pro adds: branching, merge, pull requests, access control, tags, rollback, cherry-pick, fork, multi-tenant isolation, immutable decision traces, the intelligence layer (Cortex), and a web dashboard.
OSS = single-branch repo with full history. Pro = GitHub.
git clone https://github.com/raia-live/amfs.git && cd amfs
uv pip install -e packages/core -e packages/adapters/filesystem -e packages/sdk-python -e packages/cli -e packages/http-server
uv run pytest tests/ -v- GitHub Discussions — questions, ideas, show & tell
- Roadmap — what's shipped and what's next
- Issues — bug reports and feature requests
Apache License 2.0 — free for commercial use, modification, and distribution.