chatbottool-mcp
A local-first MCP server providing guarded access to a workspace with file operations, search, commands, tests, Git helpers, checkpoints, and structured tool results. It supports multiple tool modes and emphasizes security with workspace restrictions and secret blocking.
README
<div align="center"> <img src="assets/mark.svg" width="96" alt="Chatbottool MCP mark" />
Chatbottool MCP
Give AI useful tools—not the keys to your machine.
A local-first MCP workspace bridge with guarded files, controlled execution,<br> recoverable changes, and a deliberately narrow trust boundary.
<img src="assets/chatbottool-mcp-hero.png" width="100%" alt="A luminous guarded bridge between an AI client and a local code workspace" />
The idea
Local agents are useful because they can touch real files, run real tests, and make real changes. That is also exactly what makes a careless bridge dangerous.
Chatbottool MCP puts a small, inspectable capability layer between the client and the workspace.
<table> <tr> <td width="25%" valign="top"> <h3>◇ Guarded paths</h3> Every file operation stays inside an explicitly allowed and opened workspace. </td> <td width="25%" valign="top"> <h3>⌘ Controlled work</h3> Commands, tests, Git helpers, and network access follow separate policies. </td> <td width="25%" valign="top"> <h3>↺ Recoverable change</h3> Checkpoints, trash-backed deletion, diffs, and continuity notes make work inspectable. </td> <td width="25%" valign="top"> <h3>⬡ Client-neutral</h3> Use stdio locally or HTTP + OAuth with any compatible MCP client. </td> </tr> </table>
How it works
<div align="center"> <img src="assets/architecture.svg" width="100%" alt="MCP client passes through Chatbottool policy controls before reaching a workspace" /> </div>
The client never receives ambient authority. It sees only the tools exposed by the selected mode, and each tool still passes through path, process, network, secret, and workspace checks.
| Layer | Responsibility |
|---|---|
| Workspace registry | Opens only roots already permitted by the server policy |
| Path guard | Blocks escapes, secret globs, symlink traversal, and case-confused paths |
| Tool policy | Chooses the visible capability surface: minimal, standard, or full |
| Execution layer | Runs commands, tests, Git helpers, previews, and managed processes |
| Recovery layer | Records audit events, checkpoints changes, and preserves continuity notes |
| Transport | Serves local stdio or authenticated streamable HTTP |
[!IMPORTANT] This is a capability boundary, not a sandbox against a compromised operating- system account. Run it as an unprivileged user and keep allowed roots narrow.
Two practical surfaces
| Mode | Tools | Intended use |
|---|---|---|
standard |
37 | Everyday reading, editing, structured commands, tests, safe Git writes, checkpoints, and continuity |
full |
53 | Adds 16 opt-in capabilities for trusted clients and tightly scoped workspaces |
Full mode adds full shell, recoverable deletion, background processes, repository initialization and branch recovery, Python module execution, direct file placement, and allowlisted network fetch.
The public edition deliberately contains no browser automation, desktop control, computer-use bridge, private service integration, or Git push tool.
Quick start
Requires Node.js 22.19+ (below Node 27), Git, and macOS or Linux.
git clone https://github.com/tamvi-journal/chatbottool-mcp.git
cd chatbottool-mcp
npm install
npm run build
node dist/cli.js init
The initializer creates a local policy file and a private owner password under
~/.chatbottool-mcp/.
Local MCP clients
node /absolute/path/to/chatbottool-mcp/dist/cli.js serve-stdio
{
"mcpServers": {
"chatbottool-mcp": {
"command": "node",
"args": ["/absolute/path/to/chatbottool-mcp/dist/cli.js", "serve-stdio"]
}
}
}
HTTP + OAuth
node dist/cli.js serve
The server binds to localhost by default and exposes /mcp, /healthz, and a
small status dashboard. Remote clients need a public HTTPS origin that you
operate; place the server behind a trusted reverse proxy or tunnel and set
CHATBOTTOOL_PUBLIC_BASE_URL to that origin.
Tool families
| Family | Examples |
|---|---|
| Workspace | open_workspace, ls, read, read_many, grep, glob |
| Editing | write, edit, apply_patch_many, copy_path, move_path, rename_file |
| Execution | run_command, run_test, run_build, run_pytest, run_npm_script, run_lint |
| Git | git_status, git_diff, git_add, git_commit, git_log, show_changes |
| Recovery | checkpoint_create, checkpoint_diff, checkpoint_restore |
| Continuity | continuity_save, continuity_load, handoff_note |
| Local preview | preview_local_app, collect_artifact |
Security defaults
- Common secret paths such as
.env, SSH keys, PEM files, and private keys are blocked. - Symlink escapes and case-only path confusion are rejected.
- Network fetch is full-mode only and requires an explicit hostname allowlist.
- HTTP MCP uses OAuth and host-header validation.
- Audit events redact common secret-bearing input fields.
- Shell, writes, tools, widgets, network hosts, and secret names have independent policies.
Read the complete security model and deployment checklist.
Verify
npm test
npm run typecheck
npm audit --omit=dev
node dist/cli.js doctor
The suite verifies both published surfaces: exactly 37 standard tools and 53 full-mode tools, alongside workspace confinement, OAuth, secret blocking, checkpoint recovery, audit redaction, and controlled Git behavior.
Project status
Chatbottool MCP is an early public extraction. Its security boundaries and tool
surfaces are tested, but configuration and public APIs may evolve before 1.0.
License
MIT. See LICENSE and NOTICE.md.
<div align="center"> <sub>Power where it helps. Boundaries where they matter.</sub> </div>
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.
E2B
Using MCP to run code via e2b.