mcp server
elitea-platform
The open platform for building, running and sharing AI agents, pipelines and toolkits.
Set up
Runs on your computer. Pick your agent and operating system.
This project does not publish an install method we can turn into a config. Follow the README instead.
README
synced Oct 29, 2026Elitea AI brings chat, agents, pipelines, toolkits and platform administration into one workspace your organisation runs itself. You keep the models, the credentials and the data; Elitea gives every team the same governed way to put them to work.
- Chat with models, agents, pipelines, toolkits and MCP servers, and build them inline.
- Agents and pipelines: instructions, a model and tools, or a visual workflow that chains them.
- Toolkits and MCP servers connect Jira, GitHub, Confluence and many more, with credentials kept in a vault.
- Indexing turns documents and sources into vector indexes, so answers rest on your content.
- DeepWiki and Inventory turn a code repository into browsable documentation or a queryable component graph.
- Governance: one LLM gateway with budgets, request logs and guardrails; SSO (OIDC, SAML), roles and audit.
- Your look: a runtime brand pack sets the name, logo, colours and fonts without a rebuild.
Clients
| Client | Where |
|---|---|
| Web | Served by every deployment at /app/; the admin console at /admin/app/ |
| Desktop (macOS) | A Tauri app with a local agent loop. Builds are attached to releases |
| Mobile (iOS, Android) | An Expo app that connects to a deployment by QR code (apps/elitea-mobile) |
Get started
| I want to… | Start here |
|---|---|
| Try Elitea on one machine | Install Elitea |
| Run it in production | Production deployment overview, then Install with Helm |
| Work on the code | Set up a development environment |
| Use or build on the brand | Brand guide (llms.txt for agents) |
The whole platform on one machine, from a clone, with podman:
deploy/scripts/standalone-stack.sh certs # once: local mTLS material
deploy/scripts/standalone-stack.sh build # images from source
deploy/scripts/standalone-stack.sh up # migrations, roles and the worker's identity included
open http://localhost:8084/app/
Run the standalone stack
covers seeding a real model, the overlays and the check command.
What's in this repository
| Path | What's there |
|---|---|
services/ | Go: elitea-main (API and browser edge), elitea-scheduler, elitea-llm-gateway, elitea-subapp-host, elitea-egress-proxy. Rust: the agent worker and the DeepWiki, Inventory, vector, code-runner and document-extractor engines |
apps/ | elitea-web (web app, admin console, embedded docs), elitea-desktop, elitea-mobile |
libs/ | Shared Go, Rust and TypeScript libraries and the protobuf contracts |
deploy/ | Compose stacks, the Helm chart (oci://ghcr.io/elitea-ng/charts/elitea), Argo CD and kind recipes |
brand/ | The brand guide's sources (BRAND.md) |
Architecture rules for contributors and coding agents are in AGENTS.md.
Images and the chart are published to ghcr.io/elitea-ng, signed with cosign, and
carry signed SLSA build provenance and SBOMs
(verify a release).
The charts are listed on Artifact Hub.
Contributing
Read CONTRIBUTING.md and the Code of Conduct. Report security issues privately, as described in SECURITY.md.