Skip to content

Failed to execute any MCP tool #28

@meterglost

Description

@meterglost

Context

  • I want to integrate FastCode MCP with Antigravity
  • I have setup and run FastCode using docker
  • I tried to use to sse mode but failed, it seem like due to modelcontextprotocol python sdk, not by FastCode
  • So instead I setup to run MCP using docker exec and able to get the FastCode MCP tool list
  • However the agent always get error return when try to use any tool

Input

When try to use fastcode / code_qa agent generate an input like this

{
  "multi_turn": true,
  "question": "Generate a comprehensive \"Audit Context Map\" to prepare for a whitebox engagement. Focus on mapping the application's DNA, logic flows, and attack surface.\n\nAnalyze the codebase based on the following 5 dimensions:\n\n1. Tech Stack & Dependency Fingerprinting\n   -  Identify languages, frameworks (and versions if visible), and ORMs.\n   -  List critical 3rd-party libraries, specifically looking for those involved in auth, crypto, or parsing (XML, JSON, YAML).\n   -  Identify the build chain and deployment environment (Docker, Kubernetes, CI/CD configs).\n\n2. Entry Point & Attack Surface Mapping\n   -  List all external entry points: HTTP Routes/API Endpoints, CLI arguments, WebSocket handlers, or IPC listeners.\n   -  Categorize them by privilege level (Public, User, Admin, Internal-System).\n   -  Highlight endpoints that accept complex inputs (file uploads, serialized objects, raw XML).\n\n3. Security Architecture & Controls\n   -  AuthN: How are users verified? (JWT, Session, OAuth, Basic Auth). Where is the logic located?\n   -  AuthZ: How are permissions enforced? (RBAC, ABAC, Middleware, decorators).\n   -  Sanitization: What global middleware or utility functions handle XSS/SQLi prevention?\n   -  Crypto: What algorithms are used? Where are keys/secrets likely loaded from (Env vars, hardcoded, Vault)?\n\n4. Data Flow & Critical Sinks\n   -  Trace the flow of data from entry points to sinks.\n   -  Identify database interaction patterns (Raw SQL vs. ORM).\n   -  Flag \"Dangerous Sinks\" specific to the language (e.g., eval(), system(), unpickle, render_template without escaping).\n\n5. High-Value Targets (Business Logic)\n   -  Summarize the core business functionality. What is the most critical operation this app performs (e.g., Money Transfer, File Storage, User Management)?\n   -  Identify \"Complex Logic\" files—files with unusually high lines of code or cyclomatic complexity, as bugs often hide there.\n\nOutput Format:\nPresent the findings in a structured Markdown report. End with a \"Recommended Audit Focus\" list, prioritizing which controllers or modules I should review manually first.\n",
  "repos": [
    "/workspace/projects/**censored**"
  ]
}

Ouput

Error in MCP tool execution: Error executing tool code_qa: Cannot send a request, as the client has been closed.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions