livecontext-ai

LiveContext

Community livecontext-ai
Updated

The AI automation platform, self-hosted. Describe the job in chat and LiveContext builds it: readable workflows, scoped AI agents, and small apps your team uses. Chat, Workflow, Agent and App on one canvas.

LiveContext

The AI automation platform. One message in, a working automation out.

Describe the job in chat and LiveContext builds it in front of you: a workflow you can read,AI agents with scoped access and budgets you control, and a small app your team actually uses.Chat, Workflow, Agent and App in one self-hosted platform. No code to write, nothing to stitch together.

An open-source, self-hosted alternative to n8n, Zapier and Make, with AI agents built in.

GitHub starsLatest releaseDiscussionsLicense: AGPL v3Java 21Next.jsDocker ComposeSelf-hosted

The builder, built by chat: one message in, a working automation out. Five real scenarios, one loop. Watch it full size ยท Try the hosted version

โญ If LiveContext looks useful, give it a star. It helps other teams find it.

Build it once. It runs as all four.

Most teams wire together a chatbot, an automation tool, an app builder and an agent framework.LiveContext is all four on one canvas, every agent scoped, budgeted and audited, and you can seeexactly what each one did. The chat (shown above) builds it; here is what it runs as:

Workflow + App The workflow and the app it drives, in one view. Draw the automation as a readable graph, then wrap it in a real interface: forms, dashboards and live approval screens your team or an agent can act on. Agents A fleet of scoped agents, one per job: each with its own model, tools, files, credit budget and full audit trail. No black box.
Tables Built-in data tables your workflows and agents read, write and enrich. Filter, search and export, with no external database to wire up. Data & metrics Every run charted: calls, tokens, success rate and duration, sliced per agent and per tool. Spot a regression and drill straight into it.

The workflow decides exactly what each agent sees and what it ships, so the same job runs at afraction of the cost of a do-everything agent, every step is auditable, and your business neversits inside a black box.

This repository is the Community Edition (CE): the full platform as a single self-hosted service(see LICENSE). It is free to self-host and use in production inside your organization.

Requirements

  • Docker Engine 24+ with Compose v2 (or Docker Desktop 4.x and later)
  • 4 GB RAM minimum, 8 GB recommended

Quick start

The fastest way is one npm command (Docker must be installed and running):

npx livecontext

It pulls the images, boots the whole stack, and serves on http://localhost:3000.npx livecontext down stops it and npx livecontext update upgrades it. The CLI wrapsDocker Compose, it does not replace Docker.

Or run Docker Compose directly from a clone of this repo:

# From the repo root:
docker compose up -d

# Watch it come up. The "livecontext" service runs database migrations and registers
# its tools on first boot; wait until it reports "healthy" and "frontend" is up:
docker compose ps

Then open http://localhost:3000 and create the first account (the first user becomes the admin).Two optional add-ons (interface screenshots/PDFs, and a browser agent with web search) are one envfile away when you want them, see Optional features below.

Configuration (LLM keys, SMTP, ports) is documented in docker/README-CE.md.Copy docker/.env.ce.example to set your own values, and never commit it.

Optional features

Two heavy features are opt-in and start with no container by default, keeping the base stacklight. Each is enabled by a bundled env file (it turns on both the Docker profile and the matchingapp setting in one shot):

  • Interface screenshots and PDFs (renderer profile). Adds a headless Playwright/Chromiumsidecar (~1 GB image) so interface nodes can render a PNG screenshot or a PDF. Enable it with:
    docker compose --env-file docker/.env.ce.renderer up -d
    
  • Browser agent and web search (browser-agent profile). Adds a Chromium browser-use containerplus a SearXNG metasearch sidecar (~2 GB) so agents can browse pages (agent_browse) and runweb_search. Enable it with:
    docker compose --env-file docker/.env.ce.browser-agent up -d
    

Run both by passing both env files (repeat --env-file). See docker/README-CE.mdfor details and tuning.

What's in the box

  • Workflow engine. Visual builder and execution engine with parallel branches, loops, signals,human-approval steps, and triggers (schedule, webhook, chat, form, datasource).
  • AI agents. Chat agents that design, build and run workflows, with per-workspace skills, scopedtool access, per-agent credit budgets and per-agent metrics.
  • Integration catalog. 600+ ready-made integrations seeded at first boot, fully offline. Add yourown as OpenAPI specs.
  • Interfaces and apps. Small web pages served by your workflows (forms, dashboards, approvalscreens), shareable as standalone apps.
  • Tables. Built-in data tables your workflows and agents can read and write.
  • One backend. All backend services run as a single monolith JAR, with PostgreSQL, Redis, anS3-compatible object store and a lightweight tools bridge as its dependencies, plus the Next.jsfrontend. It all comes up with one docker compose up.

Why self-host LiveContext

  • You stay in control. Per-agent credit budgets, scoped access, a full audit trail and per-agentmetrics. No black box.
  • Far fewer tokens. The workflow constrains exactly what each agent sees and ships, so jobs cost afraction of a do-everything agent.
  • Org-grade access. Organizations and workspaces with role-based access control.
  • Yours to run. The same platform on your own infrastructure.

Managed version

Prefer not to run your own infrastructure? The managed service, with an always-current integrationcatalog and hosted account management, lives at livecontext.ai. Thosehosted-only features are not part of the Community Edition.

Building from source

CE runs from prebuilt images (the Quick start above pulls them). The full source is in this repo.To build the images yourself instead of pulling, use the per-service Dockerfiles:backend/monolith-service/Dockerfile (Java 21, the ce Maven profile), frontend/Dockerfile(Node 20), and mcp/bridge/Dockerfile.

Security

Please report vulnerabilities privately. See SECURITY.md.

License

LiveContext CE is licensed under the GNU Affero General Public License v3.0(AGPL-3.0), see LICENSE. You are free to use, self-host, modify andredistribute it, including commercially. One condition matters most: if you run amodified version as a network service, the AGPL requires you to make thecorresponding source of your changes available to that service's users.

The LiveContext name and logo are trademarks of their owner and are notcovered by the AGPL, see TRADEMARKS. Third-party components shipunder their own licenses, see NOTICE andTHIRD_PARTY_NOTICES.

โญ If LiveContext is useful to you, star the repo. It is the simplest way to help other teamsdiscover it, and it means a lot to a small team. Questions or ideas? Open aDiscussion.

MCP Server ยท Populars

MCP Server ยท New

    livecontext-ai

    LiveContext

    The AI automation platform, self-hosted. Describe the job in chat and LiveContext builds it: readable workflows, scoped AI agents, and small apps your team uses. Chat, Workflow, Agent and App on one canvas.

    Community livecontext-ai
    timescale

    rsigma

    A complete Sigma detection engineering toolkit: parser, linter, evaluator, correlation engine, conversion framework, streaming daemon, MCP and LSP servers :crab:

    Community timescale
    Agent360dk

    Browser MCP by Agent360

    Drive your real, logged-in Chrome from any AI agent (Claude Code, Cursor, VS Code) โ€” works where headless dies. Reads emailed login codes from your Gmail, solves CAPTCHAs, 34 tools. MIT, local-only.

    Community Agent360dk
    devlint

    GitWand

    The Git client that actually resolves conflicts โ€” 10 deterministic patterns auto-resolve the trivial 95%, full trace on the rest. Native (Tauri 2 + Rust), free, MIT.

    Community devlint
    BETAER-08

    amdb

    Turn your codebase into AI context โ€” entirely on your machine. Single-binary MCP server with AST parsing, call graph, and local embeddings.

    Community BETAER-08