Skip to content
MCP server

Predictive Debugger

Predicts where JavaScript and TypeScript code will fail at runtime, for coding agents

Runs locallyStatus and history
Install
Pricing
Open source
Platforms
CLI, self-hosted
Runs
Locally
Version
0.9.0
Listing updated
1 day ago

About Predictive Debugger

Predictive Debugger is an open-source MCP server for developer tools. It runs locally on your machine and is started by your AI client.

Local analysis narrows the search before any model is called. One tool ranks files by risk density, another reports complexity and async-boundary signals, one returns TypeScript compiler diagnostics, one traces imports, reverse imports and connected tests with source-line evidence, and one flags unusual log entries. Only the failure prediction tool calls a model, and it returns a verdict with a line number, a reason and a confidence score. The site's example is a race in which two concurrent stock reservations read the same value and overwrite each other.

Predictions go through the Claude Code, Codex or GitHub Copilot CLI already installed and signed in, so no separate API key is needed and that provider's terms and usage allowance apply. It needs Node.js 22 or later, and Python 3 is optional for log analysis.

The site reports a benchmark of 49 files with three coding agents in which the tool used 32 to 69 percent fewer tokens than direct reading, with similar bug counts; the authors say this describes that benchmark only and that risk density is not a probability of a bug. A clean verdict does not prove a feature correct.

Install

Run in a terminal

claude mcp add predictive-debugger -- npx -y predictive-debugger

Add to claude_desktop_config.json

{
  "mcpServers": {
    "predictive-debugger": {
      "command": "npx",
      "args": [
        "-y",
        "predictive-debugger"
      ]
    }
  }
}
Install in CursorOpens Cursor, which asks you to confirm. Or add it by hand:

Add to ~/.cursor/mcp.json

{
  "mcpServers": {
    "predictive-debugger": {
      "command": "npx",
      "args": [
        "-y",
        "predictive-debugger"
      ]
    }
  }
}
Install in VS CodeOpens VS Code, which asks you to confirm. Or add it by hand:

Add to .vscode/mcp.json

{
  "servers": {
    "predictive-debugger": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "predictive-debugger"
      ]
    }
  }
}

Add to ~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "predictive-debugger": {
      "command": "npx",
      "args": [
        "-y",
        "predictive-debugger"
      ]
    }
  }
}

Add to your client's MCP settings (most use this shape)

{
  "mcpServers": {
    "predictive-debugger": {
      "command": "npx",
      "args": [
        "-y",
        "predictive-debugger"
      ]
    }
  }
}

Tools this server provides (7)

Taken from the project documentation; the running server may differ.

scan_project
Ranks source files by risk density
analyze_file
Returns complexity metrics and risk signals
check_types
Returns TypeScript compiler diagnostics
map_dependencies
Finds imports, reverse imports and connected tests
analyze_logs
Ranks log anomalies
predict_failures
Gets an independent model verdict with line, reason and confidence
list_providers
Checks which agent CLIs are installed

What it does

Risk ranking
Scans a project and ranks source files by risk density with no model call.
File signals
Shows complexity, async boundaries and contributing risk signals.
Type diagnostics
Returns TypeScript compiler errors for chosen files.
Dependency mapping
Traces imports, reverse imports and connected tests.
Log anomalies
Ranks unusual log entries by severity and wording.
Model verdicts
Gives a line number, reason and confidence for likely failures.
Existing CLI use
Calls the agent CLI you already have installed.

Good for

  • Find the riskiest files in a codebase before reviewing
  • Check a service for concurrency bugs before shipping
  • See which tests connect to a changed file
  • Look for unusual entries in an application log
  • Reduce tokens an agent spends reading code

Pricing

Open source

Prices change. Check predictivedebugger.dev before you buy.

Predictive Debugger FAQ

Do I need an API key?

No. Prediction calls use the Claude Code, Codex or Copilot CLI you have installed and signed in to.

Which tools call a model?

Only the failure prediction tool; the other six run locally.

What does it require?

Node.js 22 or later, and Python 3 only for log analysis.

Is the benchmark general?

The authors say it describes their benchmark, not accuracy on every project.

Is Predictive Debugger free?

Yes. Predictive Debugger is open source, so you can use and self-host it for free.

How do I add Predictive Debugger to Claude Code?

Run this command in your terminal: claude mcp add predictive-debugger -- npx -y predictive-debugger

What are the best alternatives to Predictive Debugger?

The closest alternatives to Predictive Debugger are REA, Codebase Memory, GitHub, Serena MCP: the IDE for your agent and XERJ.

Alternatives to Predictive Debugger

  1. REA

    MCP server

    CLI and MCP server that lets coding agents reverse engineer apps and binaries

    Runs locally
    • Open source
    • 464K GitHub stars
    • Developer tools
  2. Codebase Memory

    MCP server

    Local MCP server that builds a code knowledge graph for AI coding agents

    Runs locally
    • Open source
    • 48.7K GitHub stars
    • Knowledge and memory
  3. GitHub

    MCP server

    Official MCP server connecting AI assistants to GitHub repos, issues and workflows

    Not checked yet
    • Remote
    • Open source
    • 34.5K GitHub stars
    • Developer tools
  4. MCP toolkit giving coding agents symbol-level code retrieval and editing

    Runs locally
    • Open source
    • 31.5K GitHub stars
    • Developer tools
  5. XERJ

    MCP server

    Local search engine for AI agents with cited answers over your code and docs

    Runs locally
    • Open source
    • 19.2K GitHub stars
    • Knowledge and memory
  6. Ouroboros

    MCP server

    Spec-first layer that pins requirements and verifies results of AI coding agent work

    Runs locally
    • Open source
    • 6.4K GitHub stars
    • Developer tools