Set up Synapse

• Code graph (briefs, blast radius, tests to run): TypeScript, JavaScript and Python. More languages are coming.
• The why behind code and the PR bot: any language on GitHub, from pull requests, reviews and commits.
• AI tools: Claude (claude.ai, Desktop, Cowork), Claude Code and Cursor, over MCP.

• Sign up. Every new workspace gets the Team plan free for 14 days.
• In Repositories, click Connect GitHub and install the Synapse GitHub App on the repositories you choose. It reads code and history; its only write is one optional comment per pull request.
• Pick the repositories to add. Synapse scans each one (usually a few minutes) and reads its pull request history.

• In Claude, open Settings → Connectors and add a custom connector (or pick Synapse from the directory once it's listed).
• Name it Synapse and use this URL:

https://asksynapse.in/api/v1/mcp

Click Connect, sign in to Synapse and click Allow. No token is needed. Then ask, for example: “Using Synapse, why does the refresh token have a grace window?”

Create a token in Settings → AI agents (it's shown once), then run:

claude mcp add --transport http synapse https://asksynapse.in/api/v1/mcp --header "Authorization: Bearer YOUR_TOKEN"

Optional, recommended: the Synapse hook adds the brief to every coding prompt automatically, so the agent starts at the right files. Settings → AI agents has the exact setup for your token.

Add this to .cursor/mcp.json in your project (token from Settings → AI agents):

{
  "mcpServers": {
    "synapse": {
      "url": "https://asksynapse.in/api/v1/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_TOKEN"
      }
    }
  }
}

All tools are read-only and scoped to your workspace.

• plan_change: Plan a code change: the files and line ranges to touch, their callers, the pattern to follow, decisions not to break and the tests to run. Call it once at the start of a task.
• read_code: Read one function, or a file's outline, from the latest scan.
• why: Why a file or line is the way it is: the pull requests, reviews and commits behind it, and who knows it.
• blast_radius: What depends on a function or file: every caller and importer that could break.
• find_existing: Find existing code to reuse before writing new code.
• tests_for: The tests that cover a file or function.
• repo_map: A map of the repository: areas, entry points and key files.
• show_pattern: How this repository already does something, with an example to copy.
• check_change: Check a diff before finishing: what it touches, what it may break, and which tests to run.
• list_repositories: The repositories connected to your workspace.

On every pull request in a connected repository, Synapse posts one comment: risk, what the change can break, the decisions behind the lines being changed, and the tests to run. It updates the same comment on every push. Turn it off per repository in Synapse.

No account needed: open asksynapse.in/github and paste a GitHub link, or change github.com to asksynapse.in/github in any repository's address. Synapse maps it in a minute or two (repositories up to 150 MB), then you get four tabs:

• Ask: “where is X handled?” in plain words. You get the files and lines, checked by AI against the code when available and labelled when not. Questions like “give me its architecture” get an overview.
• Overview: the main areas, entry points and most used code.
• Cleanup: code nothing uses, duplicate functions and calls inside loops.
• Files: every file, with its pull requests and the decisions behind it.

Public pages show code that is already public. For private repositories, connect GitHub in a workspace. How accurate the answers are: see benchmarks.

Free, Pro, Team and Growth plans differ in developers, repositories, briefs per month (calls to plan_change) and the PR bot. See pricing. When a month's briefs run out, your agent is told and carries on without Synapse; other tools keep working.

Your code is never used to train AI models. Synapse doesn't log your prompts or your agent's conversations. Every workspace is isolated. Details in our Privacy policy and Terms.

Questions or problems: write to support@asksynapse.in. We usually reply within a day.