Skip to content

Choosing an integration ​

AutoRAG can be driven by a person or by another agent in several ways. This page maps each integration to the case it fits, then gives the exact prompts that make a coding agent install and configure AutoRAG for you.

What each integration is ​

IntegrationWhat it isChoose it whenStart with
Lite MCP serverautorag-mcp, a stdio MCP server with no search model; it returns ranked chunksA coding agent (Claude Code, Codex, Cursor, Pi) should search your files with its own modelMCP server (Lite)
Agent skillsMarkdown skills under skills/ that teach an agent to install, configure, and repair AutoRAGThe agent must bootstrap or repair AutoRAG, not just query itAgent skills
Full librarianThe model-backed autorag CLI and TUI, with MinSync/Jikji retrieval and Jev routingYou want a curated answer (a synthesized, cited reply) rather than raw chunksQuickstart
Remote agentsautorag-mcp --http behind an HTTPS tunnel, for agents that run in a vendor cloudAn always-on cloud agent (Claude Duet, OpenAI dots, Grok Bot, Meta Muse) should reach your index by URLAlways-on cloud agents
Hermes pluginA portable Agent Plugins v1 package that runs the Lite MCP server inside HermesYou use Hermes and want retrieval plus a search paneHermes plugin

The Lite MCP server and the full librarian share the same indexing and retrieval engine. Lite adds no model and no agent loop: the calling agent plans the search and writes the answer. The full librarian adds one configured model that plans, reads, and curates the answer itself. See Retrieval for the engine both use.

Use the Lite MCP server for external-agent search. Use the skills only to install, configure, or diagnose AutoRAG, or to drive the model-backed librarian. A wording like autorag-lite-search names a removed skill: routine Lite retrieval goes through MCP tools, not a shell command or a search skill.

Prompts that make an agent install AutoRAG ​

Paste one of these into a coding agent that can read URLs. The agent fetches the matching skill from raw.githubusercontent.com and follows it, asking before it touches anything.

Set up the AutoRAG librarian (autorag-setup):

text
Install and configure AutoRAG Agent on this computer by following
https://raw.githubusercontent.com/Marker-Inc-Korea/AutoRAG/main/skills/autorag-setup/SKILL.md
Ask me which folders to index and how to sign in to a model before you change anything.
When it is set up, run a health check and a test search, and show me the result.
Never move, rename, edit, or delete my documents while setting up.

Give your coding agent model-free search over your files (autorag-lite-setup):

text
Install and register the AutoRAG Lite MCP server for this agent by following
https://raw.githubusercontent.com/Marker-Inc-Korea/AutoRAG/main/skills/autorag-lite-setup/SKILL.md
Ask me which folders to index, build the index, then verify with a real search.
Never move, rename, edit, or delete my documents.

Something broken? (autorag-doctor):

text
Check AutoRAG by following https://raw.githubusercontent.com/Marker-Inc-Korea/AutoRAG/main/skills/autorag-doctor/SKILL.md
and finish with the per-datasource status table.

For search only, after setup, the agent calls the MCP tools directly: no additional skill is needed. See Agent skills for the tool sequence.

Released under the Apache-2.0 license.