Agent Setup — LumiBase

For AI agents: This page is also available as clean Markdown. If you are reading HTML, request index.md instead. For the full page index, see llms.txt.

LumiBase provides a prompt template, MCP server configs, and structured context files so AI coding agents can understand the codebase, call the API, and build features correctly. Pick your agent below to get started.

Quick setup

Already have an agent? Paste this into any AI coding agent to load the LumiBase project context in one step:

Fetch https://raw.githubusercontent.com/khuepm/lumibase/main/docs/en/agent-setup/prompt.md

Or if you're working locally, tell your agent:

Read docs/en/agent-setup/prompt.md and follow all setup instructions.

Manual setup — Pick your agent

Select your agent to view step-by-step instructions for wiring up LumiBase context and (optionally) the MCP server.

Agent-friendly docs

LumiBase structures its documentation to be token-efficient and easy to consume in a context window.

Markdown versions

Append /index.md to any LumiBase docs URL to get a clean Markdown version.

Page indexes (llms.txt)

Every top-level section has its own llms.txt — a page index sized to fit in a single context window:

• docs/llms.txt — directory of all LumiBase docs.
• docs/en/llms.txt — full English docs index.

Agent prompt

docs/en/agent-setup/prompt.md — machine-readable setup instructions for configuring any agent to work with LumiBase. Designed to be fetched and executed directly by an agent.

What agents can do with LumiBase

Once set up, your agent can drive the entire Content OS over MCP. The standalone MCP server (@lumibase/mcp-server) now exposes ~80 tools across every domain; every destructive tool requires an explicit confirm: true:

• Read and write collections — GET/POST/PATCH/DELETE /api/v1/items/:collection
• Manage schema — create collections, add fields, configure relations via CMS API
• Administer RBAC — roles, policies, permission rows, API keys, bulk access export/import & conflict checks
• Manage users & teams — invite/update/remove members, team membership
• Govern content — content intents (SLOs), drift scans, and reconciliation
• Trigger and monitor Flows — POST /api/v1/flows/:id/run, GET /api/v1/flows/:id/runs
• Configure delivery — presets, settings, translations + translation memory, webhooks, search
• Operate the site — activity log, health, metrics, NDJSON backup/restore, materialized collections, extensions & marketplace
• Query the AI Copilot — POST /api/v1/ai/chat with natural-language instructions
• Use WebSocket realtime — subscribe to collection changes, presence, collaborative cursors

For governed, HITL/autonomy-gated execution, agents can instead call the in-process MCP endpoint POST /api/v1/mcp (enable the per-site contentOs.mcp flag).

All endpoints require a valid access token (Authorization: Bearer <token>) scoped to the site (X-Site-Id header or subdomain routing). See api/hono-api-spec.md for the full REST/WS reference.

Compare agents

Resources

• REST API spec: docs/en/api/hono-api-spec.md
• Data model: docs/en/data-model.md
• AI Copilot internals: docs/en/features/ai-copilot.md
• AI skills registry: docs/en/ai-skills.md
• Flows automation: docs/en/features/flows-automation.md
• Architecture overview: docs/en/architecture/overview.md
• Permissions / RBAC: docs/en/features/permissions-rbac.md
• Deployment: docs/en/deployment/overview.md