MCP server
Drive ronda from your AI agent
ronda exposes a Model Context Protocol server. Connect Claude, Cursor, ChatGPT, or any MCP-capable agent, and it can open a sandbox on your repo, make a change, and open a pull request your team reviews.
MCP server URL
https://mcp.useronda.comWhat it does
Your agent does the work, your team reviews the result
An MCP-capable agent connects to ronda with a key you mint, opens a sandbox session on one of your GitHub repositories, and issues plain-English edit instructions. ronda runs your real dev server in that sandbox, the internal agent writes the change against your code, and the agent can open a pull request against your default branch when the work is done. The agent never merges anything. The output is a standard GitHub pull request your engineers review, the same safety step that applies when a person uses ronda directly. ronda is one element in your agent chain rather than a place your code has to move into.
How it works
From an agent prompt to a pull request
- 1
Connect the MCP server
Add the ronda MCP server to your agent with a key you mint in Settings. The default key covers opening sandboxes, instructing edits, reading state, and opening pull requests.
- 2
The agent makes the change
Your agent opens a session on a repo, describes the change in plain English, and ronda writes it against your real code in an isolated sandbox.
- 3
It opens a pull request
When the work is done the agent opens a standard GitHub pull request against your default branch. Your engineers review and merge it.
Why this matters
One tool in your agent chain
Bring your own agent
Use Claude, Cursor, ChatGPT, or any MCP-capable client. ronda is the editor in the chain, not another product surface your team has to adopt.
A reviewed pull request
The agent proposes and your engineers approve. Every change the agent makes lands as a pull request, so nothing the agent does reaches users unreviewed.
Your repo stays the source of truth
The agent works against your real GitHub repository. There is no separate workspace to sync and no code to move out afterward.
For builders
Integration references
The full MCP reference and the agent skill file are published as plain markdown so an agent can read them directly.
Questions
ronda over MCP
Which agents can drive ronda?
Any client that speaks the Model Context Protocol. That includes Claude, Cursor, and ChatGPT, along with other MCP-capable agents. You connect the ronda MCP server with a key and the agent can open sandboxes, instruct edits, and open pull requests.
Can the agent merge or deploy on its own?
No. The agent can open a pull request, but it cannot merge or deploy. Merging and deploying happen in your existing workflow, so a human with merge rights always approves what ships.
What is the server URL and how do I authenticate?
The MCP server is at https://mcp.useronda.com. You authenticate with a bearer token you mint in Settings under MCP. The default workspace key can open sandboxes, instruct edits, read state, and open pull requests.
Is it safe to let an agent edit my app this way?
Yes, because the agent cannot reach production directly. Every change becomes a pull request your engineers review and merge. The agent also works in an isolated sandbox running your dev server, not against your live deployment.
What does it cost?
MCP access is part of ronda's plans. The Free plan is $0 and the Starter plan is $9 per month, both bring-your-own Anthropic key, so the agent's edits use your own key and you pay Anthropic directly for the tokens.