Runyte MCP
A modal Rust editor with built-in MCP: an agent reads buffers and terminals, and terminal input waits for your approval
- 112 ★ on the repository
- Last commit
- MPL-2.0
Extended access
Why this level
- The agent reads unsaved buffers and terminal output of the selected workspace, which may hold passwords and tokens.
- With edit rights granted, the agent changes open buffers, while lines for a terminal wait for your approval.
Install
- Works in
Manual install
curl -fsSL https://raw.githubusercontent.com/runyte/runyte/main/install.sh | shInstalls the Terminal edition on Linux and macOS into `~/.local/bin` without sudo and verifies the release checksum.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The agent reads unsaved buffers and terminal output of the selected workspace, which may hold passwords and tokens.
- With edit rights granted, the agent changes open buffers, while lines for a terminal wait for your approval.
This is third-party code. Review the repository files before installing.
What it does
Runyte is a fast modal text editor with terminals, Git, LSP and Tree-sitter support for 31 languages. For agents it ships a built-in MCP adapter that starts with `runyte mcp --identity agent` over stdio and needs no Python, keys or open ports. Nine tools find buffers and terminals by name or content (`find_resources`), read unsaved text (`read_buffer`) and recent terminal output (`read_terminal`). Editing works on three levels: `append_buffer` adds text in one undoable transaction, `edit_buffer` replaces a range when the revision matches, and `propose_terminal_text` sends one line for a terminal and waits for your approval. What an agent may do is decided by you in each workspace with `:mcp agent`: terminal reads, editor context reads, buffer edits, terminal proposals.
Who it is for. People who run several agents in terminals of one editor and want Claude Code or Codex to see buffers and the output of neighboring terminals.
Good fit when
- You need to find the terminal running Claude Code and read its latest output.
- An unsaved buffer with a note draft is extended through `append_buffer`, and the change undoes in one step.
- A test command is sent to a neighboring terminal, and you approve it before it is typed.
- Codex and Claude Code share one grant: both use the `agent` identity.
- The editor opens from an agent to edit a commit message: `Ctrl-g` attaches to the already running instance.
Not a fit when
- The agent must save files or send commands to a terminal itself: the adapter does not save files or press Enter.
- The editor is new to you: everything runs through key sequences, and you will have to learn them.
- You want a windowed editor on Windows: the desktop edition exists only for Linux and macOS.
- Agent rights are meant to be granted once for all projects: access is granted separately in each workspace.
Example request
Find the terminal running Claude in Runyte and read its latest output, then send a test command for a neighboring terminal.MCP
- Transport
- stdio
- Authentication
- not required
Limitations
Runyte is an editor, and its MCP adapter ships inside the `runyte` binary, with no separate package for the adapter. The agent needs the absolute path to the binary. Rights are granted per workspace with `:mcp agent` and last until the editor closes unless you choose to remember them. Automated tests run on Linux x86-64, macOS ARM64 and Windows 11, and other platforms get builds only. The window with image and PDF viewing belongs to the Desktop edition for Linux and macOS.
How to disable. Run `claude mcp remove runyte` and `codex mcp remove runyte`, then revoke access with `:mcp` in the editor: the `x` key removes the selected identity.
FAQ
How do I connect Runyte to Claude Code?
Run `claude mcp add --scope user --transport stdio runyte -- /path/to/runyte mcp --identity agent` with the absolute path to the binary. Then restart the agent session and grant rights with `:mcp agent` in the editor.
How do I connect Runyte to Codex?
Run `codex mcp add runyte -- /path/to/runyte mcp --identity agent`. `codex mcp list` and `claude mcp list` check the settings of both clients.
What can an agent do with terminals?
Read their output and send one line of text. The line reaches the terminal only after your approval, and the adapter cannot press Enter.
Do I need Python or an API key?
No. The adapter is built into the `runyte` binary and talks to the client over stdio, with no network ports.
Which MCP tools does Runyte have?
Nine: `find_resources`, `list_workspaces`, `read_buffer`, `read_terminal`, `append_buffer`, `edit_buffer`, `propose_terminal_text`, `terminal_proposal_status` and `cancel_terminal_proposal`.
Discussion
No questions or reviews yet.
Related
MCP serversA Rust binary for agents and the terminal that uses tree-sitter to find a symbol's definition, its calls and file dependencies
A Rust coding agent with a native GUI, a terminal mode, an ACP mode for Zed and an MCP server
rust-analyzer MCP: Rust code navigation for agents
r3bl-rust-analyzer-mcp-server
A Rust MCP server on standard threads: the agent gets hover info, go to definition, references, formatting and compiler diagnostics from rust-analyzer
An MCP server and CLI for code research: search and understand your project and all of GitHub with fewer tokens
An MCP server that indexes code with tree-sitter so an agent reads single symbols, not files, traces paths and weighs edits
Code graph and project memory in one Rust program: the agent gets excerpts with file and line citations
README in short
The project describes the editor: workspaces, panes and terminals, Git and plugins, and for agents a separate MCP page with a table of nine tools, per-workspace permissions and the order of resource discovery.