Repiscope MCP server

Repiscope

A periscope for your repos — see everything, touch nothing.

Repiscope is a read-only MCP server that gives your coding agent (Claude Code, Cursor, or any MCP client) awareness of the sibling repositories next to the one it's working in — without ever letting it modify them.

Why

I built a personal assistant to help me decide where my time goes: for that it needs to know the real state of every project — what changed, what stalled, what the next step is. So it has to see all my repos. But telling an agent "don't touch anything" is a request, not a guarantee. Repiscope makes it a guarantee: the server exposes zero write tools, so the agent structurally cannot modify your repos. And the same periscope helps any coding agent working in project A that needs to know how a part of project B works.

Sandboxes make this worse, not better: mine (rightly) stops the agent from even reading outside the repo it started in without asking. Widening the sandbox would grant write access too. Repiscope is the third way — the agent gets structured, read-only sight of every sibling repo through one MCP server, and the sandbox stays tight.

See it in action

You're working in project-a and need to pull data from another project's API. Instead of you digging through that codebase, the agent asks Repiscope directly:

You:    Use Repiscope to find out what kind of data I can retrieve
        from billing-api through its API.

Agent:  (calls list_projects → sees "billing-api")
        (calls project_overview "billing-api" → spots src/routes/)
        (calls search "route" project="billing-api")
        (calls read_file "billing-api" "src/routes/invoices.py")

        billing-api exposes three endpoints: /invoices (list, filterable
        by customer and date), /invoices/{id} (full invoice with line
        items), and /customers/{id}/balance (current balance). All
        responses are JSON; auth is a bearer token.

The agent explored a repo it has no ability to write to. No copy-paste, no "please don't edit anything," no risk.

Tools

How it stays fresh

Overviews are cached as markdown and refreshed lazily: on each call Repiscope compares the repo's current git commit hash against the one recorded when the overview was built. Same hash → serve the cache. Different → rebuild just that repo's overview. No cron, no daemons.

Borrowed intelligence

Repiscope has no LLM of its own — no API key, no model calls, zero cost. But it talks to LLMs all day, so it borrows them: when an overview has no fresh agent-written summary, it ends with a note asking the calling agent to write one and hand it back via store_summary. The summary then opens every future overview of that project — written by one agent, read by all the next — until the repo's next commit marks it outdated and the cycle repeats.

Security by architecture

Repiscope is built so that the safe behaviour is not a promise — it's the only behaviour possible:

Shortened. The full README is on GitHub.