# Repowise (Hosted) — MCP server card for AI agents

Ask how your codebase works. Hosted codebase intelligence for repos indexed on repowise.dev.

Registry name: `dev.repowise/repowise-hosted` · category: Developer tools · version 1.0.0 · updated 2026-11-05

## Connect
- Remote (Claude Code): `claude mcp add --transport http repowise-hosted https://api.repowise.dev/mcp`

## Facts
- Repository: https://github.com/repowise-dev/repowise (8,170 stars)
- Website: https://www.repowise.dev

Before you connect a server: read its repository, and give it only the permissions and keys it needs.

## Links
- AgentGid page: https://agentgid.com/mcp/servers/dev-repowise-repowise-hosted/


## Repowise Hosted: ask about your code from any AI chat

The online version of Repowise. You sign in at repowise.dev, pick a GitHub or GitLab project, and Repowise builds a map of the code on its own servers. Claude, Cursor or another AI assistant then connects to one web address. It asks that map how the code works, what a change may break and why it is built that way.

### Where to use it
- Ask about your code from Claude chat: Add Repowise as a connector in Claude on the web, desktop or phone and ask how a part of your project works. You sign in once in a browser window, with no key to copy.
- One shared code map for every editor: Index the project once on Repowise and point Claude Code, Cursor, VS Code or Windsurf at the same address. Nothing runs on your computer.
- Questions across several projects: A team with a front end, a back end and shared libraries connects one workspace address and asks across all of them at once.
- Record the rules your code follows: Let your assistant write down the decisions the code already follows, each with an exact quote and its commits. It can also review the suggested ones waiting for a yes or no.
- Free map of an open-source project: Maintainers of a public project get it indexed for free and can let contributors' assistants ask about the code.
- A new teammate's first week: Give a new developer's assistant the team's hosted address so it can answer "where is this handled" without waiting for a senior colleague. (idea, not from the official pages)

### How to start
- Sign in at repowise.dev and index a GitHub or GitLab project.
- Open Settings, Coding agents (MCP), generate a key named after your editor and copy the ready snippet.
- Add the server in your editor with the project's address and the key, then restart it.

```
claude mcp add --transport http repowise https://api.repowise.dev/mcp/OWNER/REPO \
  --header "Authorization: Bearer YOUR_API_KEY"
```


### License
A paid online service with a free plan for two public projects. Private projects need the Pro plan or higher; the engine behind it is open source under the AGPL.

### Good to know
- You need a Repowise account and a key from its settings page; Claude chat signs in through a browser window instead.
- Your code is copied to Repowise's servers for each index and deleted after the run; the map built from it is kept.
- On the Pro plan, assistant calls are paid from the same monthly AI credits as chat and written docs.
- The workspace address only reads and needs a Pro, Teams or Admin account; writing tools work only on a single project address.
- If your code must not leave your own network, use the self-hosted version instead.

Checked 2026-11-10 against: https://registry.modelcontextprotocol.io/v0.1/servers?search=dev.repowise/repowise-hosted&version=latest, https://github.com/repowise-dev/repowise/blob/db12feee74bbff347ea66344c23c113517ed21f6/README.md, https://github.com/repowise-dev/repowise/blob/db12feee74bbff347ea66344c23c113517ed21f6/server.json, https://github.com/repowise-dev/repowise/blob/db12feee74bbff347ea66344c23c113517ed21f6/docs/agent/MCP_TOOLS.md, https://github.com/repowise-dev/repowise/blob/db12feee74bbff347ea66344c23c113517ed21f6/packages/server/src/repowise/server/mcp_server/_http_auth.py, https://docs.repowise.dev/llms.txt, https://docs.repowise.dev/hosted/overview.md, https://docs.repowise.dev/hosted/connecting-ide.md, https://docs.repowise.dev/hosted/vs-self-hosted.md, https://docs.repowise.dev/hosted/data-handling.md, https://docs.repowise.dev/hosted/agent-decisions.md, https://docs.repowise.dev/mcp/overview.md, https://repowise.dev/pricing.md

### Mistakes to avoid
- Don't: Use the registry address https://api.repowise.dev/mcp as is → Do: Use https://api.repowise.dev/mcp/OWNER/REPO for one project, or https://api.repowise.dev/mcp/w/{workspace_id} for a workspace (The owner and repo segments are required; the bare address is only the base.)
- Don't: Put the Repowise URL into claude_desktop_config.json for Claude Desktop → Do: Add it as a custom connector under Customize, Connectors and sign in through the browser (That file only configures local servers; Claude apps connect to remote servers over OAuth.)
- Don't: Paste the Cursor mcpServers block into .vscode/mcp.json → Do: In VS Code use the top-level key servers with "type":"http"; Windsurf uses serverUrl (Each editor reads a different config shape.)
- Don't: Call decision on a workspace address or with a read-only key → Do: Turn on decision in the project's Settings, MCP tools, and connect the per-project address with a write-scoped key (Workspace endpoints are read-only and decision needs the write scope.)
- Don't: Call get_symbol on a workspace without naming a project → Do: Pass repo with an alias from list_repos (A symbol id only means something inside one project.)

### Questions
- Do I need to install anything? No. Repowise indexes the project on its servers; your editor only needs the address and a key.
- Is it the same as the free self-hosted Repowise? The tools and engine are the same; hosted adds managed indexing, team features and a few hosted-only pages and tools.
- Can the assistant change anything? Only through the decision tool, which is off until you turn it on and needs a key with write access.
- Which AI providers see my code? Only for model features such as written docs, chat and answers, excerpts go to Anthropic, OpenAI or Google Gemini models.

---
AgentGid's catalog is curated: every entry was checked against the maker's official pages (registry, repository, license and pricing pages) on the date shown; a license value shows the quote it stands on wherever AgentGid has one. Usage numbers come from AgentGid's own scan of public repositories set up for AI coding agents. Claims by makers are reported, not tested by AgentGid.
