candor
Map which functions reach network, files, database or subprocess, then gate risky edits
- Pricing
- Open source
- Platforms
- CLI, macOS, Linux
- Runs
- Locally
- Version
- 0.42.1
- Listing updated
- 3 days ago
About candor
candor is an open-source MCP server for developer tools. It runs locally on your machine and is started by your AI client.
candor builds a precomputed map of what every function can reach, traced through call chains: network, filesystem, database, clock, subprocess and similar effects. Where it cannot resolve a callee, for example a callback or an any-typed call, it reports a disclosed Unknown rather than treating the code as pure. An agent asks questions such as the blast radius of a change or what reaches the network and gets a deterministic answer from that map instead of tracing call graphs by hand.
Once you declare a boundary, the map becomes a gate: an edit that crosses it fails the build, and a baseline guard fails a scan when an existing function gains an effect. A policy can also be applied to a dependency's published report without its source, which the site pitches for catching supply-chain changes.
candor scan picks an engine from the manifest. Rust, JVM, Swift and TypeScript engines exist, and candor mcp reads whichever report is in .candor/. This registry entry is the npm candor-ts package. It can be locked to one workspace with CANDOR_ROOT.
Install
Run in a terminal
claude mcp add candor -- npx -y candor-tsAdd to claude_desktop_config.json
{
"mcpServers": {
"candor": {
"command": "npx",
"args": [
"-y",
"candor-ts"
]
}
}
}Add to ~/.cursor/mcp.json
{
"mcpServers": {
"candor": {
"command": "npx",
"args": [
"-y",
"candor-ts"
]
}
}
}Add to .vscode/mcp.json
{
"servers": {
"candor": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"candor-ts"
]
}
}
}Add to ~/.codeium/windsurf/mcp_config.json
{
"mcpServers": {
"candor": {
"command": "npx",
"args": [
"-y",
"candor-ts"
]
}
}
}Add to your client's MCP settings (most use this shape)
{
"mcpServers": {
"candor": {
"command": "npx",
"args": [
"-y",
"candor-ts"
]
}
}
}Configuration
CANDOR_ROOTLock the server to one workspace: every report prefix is resolved under this root. Unset, the server reads .candor/ relative to the client's working directory.
What it does
- Effect map per function
- Shows which functions reach Net, Fs, Db, Exec, Clock and more, transitively.
- Blast radius queries
- Lists the transitive callers of a function before it is changed.
- Policy gate
- Fails the build when an edit crosses a declared boundary.
- Baseline guard
- Fails a scan if an existing function gained an effect.
- Disclosed unknowns
- Unresolved calls are reported, never silently treated as pure.
- Dependency diff
- Applies a policy to a dependency's published report without its source.
Good for
- Ask an agent what reaches the database before refactoring
- Enforce that a domain layer does no I/O in CI
- Flag a dependency release that adds network access
- Rank the blind spots where analysis could not resolve a call
Pricing
Open source
Install via Homebrew or npm; licence not published
Prices change. Check candor.poly.io before you buy.
candor FAQ
Which languages does it cover?
The site lists Rust, the JVM, Swift and TypeScript engines.
Does it need a network call to work?
The site says the setup snippet comes from the copy on your machine, and the scan writes the map to .candor/report.json.
How do I install it?
`brew install tombaldwin/tap/candor`, then `candor scan .` and `candor mcp install`.
How do I limit it to one project?
Set CANDOR_ROOT to lock the server to a workspace.
Is candor free?
Yes. candor is open source, so you can use and self-host it for free.
How do I add candor to Claude Code?
Run this command in your terminal: claude mcp add candor -- npx -y candor-ts
What are the best alternatives to candor?
The closest alternatives to candor are REA, Codebase Memory, GitHub, Serena MCP: the IDE for your agent and XERJ.
Alternatives to candor

REA
MCP serverCLI and MCP server that lets coding agents reverse engineer apps and binaries
Runs locally- Open source
- 464K GitHub stars
- Developer tools

Codebase Memory
MCP serverLocal MCP server that builds a code knowledge graph for AI coding agents
Runs locally- Open source
- 48.7K GitHub stars
- Knowledge and memory

GitHub
MCP serverOfficial MCP server connecting AI assistants to GitHub repos, issues and workflows
Not checked yet- Remote
- Open source
- 34.5K GitHub stars
- Developer tools

Serena MCP: the IDE for your agent
MCP serverMCP toolkit giving coding agents symbol-level code retrieval and editing
Runs locally- Open source
- 31.5K GitHub stars
- Developer tools

XERJ
MCP serverLocal search engine for AI agents with cited answers over your code and docs
Runs locally- Open source
- 19.2K GitHub stars
- Knowledge and memory

Ouroboros
MCP serverSpec-first layer that pins requirements and verifies results of AI coding agent work
Runs locally- Open source
- 6.4K GitHub stars
- Developer tools