CodeWeave MCP
Provides tools to inspect, edit, and validate a Git repository, including context fetching, file reading, patch preparation and application, and running allow-listed tasks.
Allows connection to ChatGPT for AI-assisted coding.
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@CodeWeave MCPFind the definition of the User model"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
CodeWeave
CodeWeave is a local-first MCP server that gives an AI coding client precise search, guarded edits, Git operations, semantic intelligence, and bounded Bash execution inside one repository.
It is built for developers who want capable coding tools without silently switching repositories, hiding partial edits, or sending an entire machine to a remote service.
Quick start
You need Rust, Git, and Bash. Windows users can use the Bash bundled with Git for Windows.
cargo install --git https://github.com/abhij1306/codeweave --locked
codeweave installThe interactive installer:
asks which repository CodeWeave should serve;
recommends local stdio or configures loopback HTTP;
creates
config.jsonand a private origin token;checks Git, Bash, the workspace, configuration, token permissions, port, and eager index startup;
prints a ready-to-paste MCP client command.
For automation, skip the wizard:
codeweave init --path /absolute/path/to/repository --config config.json
codeweave doctor --config config.jsonRelated MCP server: MatterAI MCP Server
Connect a client
Local clients: stdio
Stdio is the simplest and safest transport. The installer prints a block like:
{
"command": "/absolute/path/to/codeweave",
"args": ["serve", "--transport", "stdio", "--config", "/absolute/path/to/config.json"]
}Paste it into any MCP client that accepts a local command.
Loopback HTTP
codeweave serve --transport http --config config.jsonThe local endpoint is http://127.0.0.1:8813/mcp by default. HTTP requires the
bearer stored in .mcp-token; the token is a private origin credential, not an
LLM API key and not external user authentication.
Remote clients
Never publish the loopback origin by simply injecting its bearer into every public request. A public gateway must authenticate each external caller first, then add the private origin bearer for the hop to CodeWeave.
The bundled PowerShell helper uses mandatory HTTP Basic authentication at ngrok before it substitutes the origin bearer:
.\start-ngrok.ps1 -Config .\config.jsonPowerShell prompts for the public username and password. Use this helper only with clients that support Basic credentials. URL-only hosted connectors need an OAuth-capable MCP gateway instead. See ChatGPT setup and Claude setup.
What CodeWeave exposes
CodeWeave advertises one fixed 25-tool surface:
discovery:
workspace,code_retrieve,code_intelligence;guarded editing:
code_write,code_replace,code_replace_range,code_insert,code_delete,code_rename,code_preview,code_transaction;Git:
git_status,git_diff,git_log,git_show,git_blame,git_preflight,git_stage,git_commit,git_restore,git_push;commands:
bash,bash_status,bash_output,bash_cancel.
The repository is fixed for the process lifetime. All connected clients share the same workspace mutations, generation, and Bash runs.
Safety model
CodeWeave is privileged developer tooling. An authenticated client can run unsandboxed Bash as your OS account, so connect only clients and models you trust.
The server still enforces narrower invariants around that trust:
file reads and writes are capability-confined to the configured workspace;
snapshot hashes, range handles, and expected-content checks guard edits;
multi-file edits report partial application and recovery failures honestly;
Git path arguments and destructive operations use validation and explicit confirmation;
HTTP authentication happens before a bounded 4 MiB request-body read;
the origin accepts HTTP/1 only and caps concurrent connections;
bearer files use
0600on Unix and a protected explicit DACL on Windows.
Read SECURITY.md before exposing HTTP beyond loopback.
Configuration
codeweave install creates a strict configVersion: 2 configuration. Unknown
fields are rejected, including unknown nested fields. See
config.example.json for every option.
Useful commands:
codeweave doctor --config config.json
codeweave serve --transport stdio --config config.json
codeweave serve --transport http --config config.jsonOptional semantic backends are rust-analyzer, basedpyright-langserver, and
typescript-language-server. They are disabled until enabled in the generated
configuration.
Do not commit config.json, .mcp-token, or .codeweave-cache/.
Updating
Install the newest release from GitHub:
cargo install --git https://github.com/abhij1306/codeweave --locked --force
codeweave doctor --config /path/to/config.jsonOlder Windows releases created bearer files with inherited permissions. If
doctor rejects such a file, stop CodeWeave, remove only the configured token
file, and start it again so the token is rotated with a protected DACL.
Developing
git clone https://github.com/abhij1306/codeweave.git
cd codeweave
cargo fmt --check
cargo clippy --all-targets --all-features -- -D warnings
cargo test --all-targetsArchitecture and concurrency guarantees live in docs/architecture.md. Tool contracts are documented in docs/tools.md, and contributions are welcome through CONTRIBUTING.md.
This server cannot be installed
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Related MCP Servers
- AlicenseBqualityDmaintenanceEnables AI assistants to perform code reviews by providing access to staged files, git diffs, and repository file content. It allows users to evaluate changes and context within any local git repository before committing or pushing.39ISC

MatterAI MCP Serverofficial
AlicenseNot gradedqualityCmaintenanceEnables code reviews, implementation planning, and pull request generation for AI agents in IDEs like Cursor and Windsurf.1MIT- AlicenseNot gradedqualityCmaintenanceProvides AI assistants with 28 developer tools across file, git, code analysis, HTTP, and system domains, enabling tasks like file editing, repository management, code analysis, and shell command execution.222MIT
- AlicenseNot gradedqualityCmaintenanceProvides AI coding assistants with tools to run git status, recent commits, tests, and linter on a real repository.25MIT
Related MCP Connectors
Git-backed platform for skills, tools, and context for AI agents
Connect AI assistants to GitHub - manage repos, issues, PRs, and workflows through natural language.
Codebase intelligence for agents: 152 structured artifacts across 21 programs, one call.
Latest Blog Posts
- Who's Calling? MCP Hosts Are an Identity Blind Spot (And the Spec Knows It)By Om-Shree-0709 on .mcpAgent IdentityOAuth 2.1
- Your AI Chatbot Just Exposed Your CEO's Salary to an InternBy Om-Shree-0709 on .Agent IdentityMCP SecurityOAuth Delegation
- Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)By Om-Shree-0709 on .Agentic AiPrompt InjectionWebAssembly
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/abhij1306/codeweave'
If you have feedback or need assistance with the MCP directory API, please join our Discord server