Skip to main content
Glama

ContextOS

CI Release License: MIT

Shared brain for AI agents.

Stop re-explaining the same codebase to every agent.

Collega tutti i tuoi agenti AI a una memoria unica, sempre aggiornata, controllabile e sicura.

ContextOS is a local-first memory layer for AI coding agents. Cursor, Claude Code, and humans share versioned decisions, conventions, tasks, conflicts, and handoffs — readable in a workspace and by machines over REST + MCP.

This is v1.0 Beta. MIT. Free on your machine. There is no ContextOS cloud, billing, or fake Pro plan.

Landing

More captures from Playwright: login · Agent Hub · workspace · memory after scan · agent brief · privacy & team · viewer read-only · pricing

10-minute quickstart

You need Node 22+, npm, and PostgreSQL. Docker is the usual Postgres; see Quickstart if Docker Desktop is not running.

git clone https://github.com/carlogiovannifalocco-byte/contextos-v1
cd contextos-v1
node scripts/setup.mjs
npm run dev

Open http://127.0.0.1:5173 (or the port Vite prints).

Seeded demo (after setup):

  • email: demo@contextos.dev

  • password: DemoPassw0rd!

  • project: Atlas CLI — two agents (Forge, Scribe) and one open conflict on ignore-file syntax

Then:

  1. Sign in and open Atlas CLI

  2. Open Brief — see the exact markdown your agents receive (budget + focus)

  3. Memory, tasks, and the conflict banner should already have a story

  4. Wire an agent — install the CLI, then init:

# from GitHub release (no npm account needed)
npm install -g https://github.com/carlogiovannifalocco-byte/contextos-v1/releases/download/v1.0.0-beta/contextos-memory-1.0.0-beta.tgz

# or when published: npx contextos-memory init
contextos init --api http://127.0.0.1:3010

See docs/MCP.md for MCP env vars. 5. Watch Activity while an agent writes memory (SSE)

Related MCP server: 0f

Tests

npm test from the repo root runs unit tests (shared schemas, env/CSP/scan-path) plus API integration tests (auth, memory, scan).

  • After node scripts/setup.mjs, Vitest loads .env automatically. Integration tests need Postgres via DATABASE_URL.

  • CI starts Postgres 16 and sets DATABASE_URL in .github/workflows/ci.yml.

  • Without DATABASE_URL (and without a .env), integration tests are skipped with one message telling you to run setup. Unit tests still pass. You should not see a pile of Prisma Environment variable not found traces.

npm test
npm run test:e2e

E2E defaults to Vite 5173 + API 3001. If those ports are taken:

$env:PLAYWRIGHT_BASE_URL="http://127.0.0.1:5174"
$env:CONTEXTOS_API_ORIGIN="http://127.0.0.1:3010"
npm run test:e2e

Production compose

Copy .env.example to .env and set a 32+ character COOKIE_SECRET that is not a placeholder. The API refuses to boot without it.

docker compose -f docker-compose.prod.yml up --build

Migrations run on API start (prisma migrate deploy via scripts/prod-start.mjs), not db push. Browser mutations require a session + CSRF. Agents use Authorization: Bearer cos_….

Verified on WSL2 Docker (Ubuntu 26.04, Engine 29.1.3). Windows Docker Desktop’s service remains Stopped on this host — use WSL or fix Desktop for native Windows compose.

What it is (and is not)

Is

Is not

Shared, permissioned memory for agents

Jira / Notion / Git

Local-first infrastructure

Cloud SaaS that needs our account

Human verify/pin + conflict merge

Unattended “AI dashboard”

Docs

Known limitations (honest)

  • Single-process realtime (in-memory SSE/WebSocket). No Redis fan-out.

  • No encryption at rest. Protect the disk and Postgres yourself.

  • No Kubernetes/Helm charts.

  • Conflict detection is lexical (similar titles on decisions/conventions/constraints), not semantic embeddings.

  • Folder scan is heuristic (package.json, README, EditorConfig) — not an LLM.

  • Organizations, memory graphs, embeddings, and a desktop wrapper are out of this beta.

  • Production Docker Compose verified on WSL2 (see docs/QUICKSTART.md). Windows Docker Desktop on this host is still Stopped.

License

MIT

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    Related MCP Connectors

    Related MCP Servers

    • F
      license
      Not graded
      quality
      D
      maintenance
      A local-first MCP server that manages developer memory for coding agents, enabling shared project context, permissions, and audit trails across different agents.
      1
      -
    • A
      license
      Not graded
      quality
      B
      maintenance
      Enables AI coding agents to share a persistent local-first memory hub, storing and recalling architectural decisions and context across different tools via MCP, so users can switch assistants without losing context.
      Apache 2.0
    • A
      license
      Not graded
      quality
      B
      maintenance
      Enables AI coding agents and assistants to share a local-first, token-efficient persistent memory layer across MCP-compatible runtimes, storing typed project facts, decisions, failures, and solutions with temporal validity. It supports offline use and Postgres-compatible storage, so agents can retrieve relevant context without re-embedding entire repositories.
      Apache 2.0
    • A
      license
      Not graded
      quality
      B
      maintenance
      Enables teams of AI coding agents (Claude Code, Cursor, Codex) to share persistent, time-aware graph memory and coordinate their work via MCP, so agents recall verified fixes at the moment of failure, learn procedures automatically, avoid file conflicts, claim tasks race-free, and hand off facts and decisions between sessions.
      1
      MIT