Skip to main content
Experimental API - Context Connectors is experimental and subject to breaking changes.

Prerequisites

  • Node.js 18+
  • Augment API credentials (AUGMENT_API_TOKEN, AUGMENT_API_URL)
  • Git provider token with repo read access:
    • GitHub: GITHUB_TOKEN
    • GitLab: GITLAB_TOKEN
    • BitBucket: BITBUCKET_TOKEN
  • For chat: An LLM API key (OpenAI, Anthropic, or Google)

Steps

1. Install

npm install @augmentcode/context-connectors

2. Set credentials

export AUGMENT_API_TOKEN='your-token'
export AUGMENT_API_URL='https://your-tenant.api.augmentcode.com/'
export GITHUB_TOKEN='your-github-token'

3. Index the repository

npx ctxc index github --owner facebook --repo react -i react
You should see:
Fetching file tree from facebook/react...
Indexing complete: 2847 files indexed, 156 skipped
npx ctxc search "How does the reconciliation algorithm work?" -i react
You should see an LLM-generated answer based on the codebase:
Answer:

Based on the code in packages/react-reconciler/src/ReactFiberReconciler.js...
For raw search results without LLM processing, add --raw:
npx ctxc search "reconciliation" -i react --raw

5. Chat (Optional)

For an interactive AI agent that can search and read the codebase:
export OPENAI_API_KEY='your-openai-key'
npx ctxc agent -i react --provider openai
You should see:
Agent ready. Type your question or 'exit' to quit.

>
Ask questions interactively:
> How does React handle component updates?

Searching: "component updates"...
Reading: packages/react-reconciler/src/ReactFiberWorkLoop.js...

Based on the code, React handles component updates by...
Type exit to quit.

Done!

You can now:
  • Search your codebase semantically with the search command
  • Chat with an AI agent that understands your code using the agent command
No need to clone the repository locally - Context Connectors fetches files directly from the Git provider API.

Also Works With

Instead of…Try…
Default branch--ref main, --ref v1.0.0, or --ref abc123 for specific branch/tag/commit
Local storage--store s3 with CC_S3_BUCKET for team sharing
Manual updatesSet up GitHub Actions to re-index on every push
Interactive chatnpx ctxc agent -i react --provider openai "your question" --print for single-question mode
OpenAI--provider anthropic or --provider google for other LLM providers
Code reposIndex websites to chat with documentation sites