Start Here · Recipe

Getting started

Create a workspace, connect a repository, and ask your first question.

This is the shortest path from nothing to a project whose knowledge you can query. It takes about ten minutes and assumes you can access a Git repository.

By the end you will have a workspace, a connected repository that Chiron has indexed, and a first answer pulled from the Ontology Layer.

1. Create a workspace and a project

A workspace groups your projects, members, billing, and default model choice. A project is one bounded effort with a repository attached. Create both from Manager, then open the project shell.

2. Connect a repository

Point the project at a GitHub or Azure DevOps repository. Chiron detects the provider from the URL and begins ingestion. Everything gets read; you separately choose which repositories agents are allowed to push to.

text

Provider is inferred from the URL host:
  github.com            -> github
  dev.azure.com         -> azure_devops
  gitlab.com            -> gitlab

The project's setup_status moves pending -> analyzing -> ready. While it analyzes, the code, graph, and docs sources fill in. See Connect a repository for the detailed walkthrough.

3. Ask the knowledge layer a question

Once the project is ready, discover what it knows and run a search. The quickest path is the MCP server (below); the raw HTTP API is documented in the Knowledge API.

bash

curl -X POST "$ONTOLOGY_URL/projects/$PROJECT_ID/knowledge/search" \
  -H "X-API-Key: $INTERNAL_SERVICE_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "query": "where do we validate a session token?", "top_k": 5 }'

You will get back a ranked list of hits, each with a ref you can resolve to full content with fetch.

4. Connect an agent

Wire the MCP server into your editor so an agent can call the same tools you just used. See Wire up MCP in your editor.

Next: First project walkthrough.