contextos-memory
This server does not currently have a deployable release on Glama. Follow the repository's setup instructions to install and configure it for your MCP client, including any required credentials and local dependencies.
ContextOS
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.

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 devOpen http://127.0.0.1:5173 (or the port Vite prints).
Seeded demo (after setup):
email:
demo@contextos.devpassword:
DemoPassw0rd!project: Atlas CLI — two agents (Forge, Scribe) and one open conflict on ignore-file syntax
Then:
Sign in and open Atlas CLI
Open Brief — see the exact markdown your agents receive (budget + focus)
Memory, tasks, and the conflict banner should already have a story
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://127.0.0.1:3010See 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.envautomatically. Integration tests need Postgres viaDATABASE_URL.CI starts Postgres 16 and sets
DATABASE_URLin.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 PrismaEnvironment variable not foundtraces.
npm test
npm run test:e2eE2E 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:e2eProduction 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 --buildAPI (MCP): http://localhost:3001
OpenAPI: http://localhost:8080/api/docs · JSON spec
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
CI integration — pull
contextos briefinto GitHub Actions
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
This server cannot be deployed
Maintenance
Related MCP Connectors
Team memory for AI coding agents and code reviews, reviewed by humans and served over MCP.
Private-by-default, local-first memory/context/task orchestrator for MCP apps and agents.
Persistent, inspectable memory for AI agents with lineage, correction, and a hosted MCP endpoint.
Hosted memory for AI agents over MCP. One memory shared by all your agents.
Related MCP Servers
- FlicenseNot gradedqualityDmaintenanceA local-first MCP server that manages developer memory for coding agents, enabling shared project context, permissions, and audit trails across different agents.1-
- AlicenseNot gradedqualityBmaintenanceEnables 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
- AlicenseNot gradedqualityBmaintenanceEnables 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
- AlicenseNot gradedqualityAmaintenanceEnables 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.MIT