Skip to content
MCP server

Trustabl Probe

Drafts a network allow list for an agent tool and test-runs it with the network blocked

Runs locallyStatus and history
Install
Pricing
Open source
Platforms
CLI, self-hosted
Runs
Locally
Version
0.5.0
Listing updated
2 days ago

About Trustabl Probe

Trustabl Probe is an open-source MCP server for security. It runs locally on your machine and is started by your AI client.

Probe checks a tool directory before an agent is trusted with it. A static pass reads the code for missing timeouts, hardcoded credentials and unfiltered tool output, and drafts a network allow list as an OpenShell network policy. An optional sandbox pass then runs the tool with its network blocked to catch crashes, missing dependencies and blocked connection attempts.

The sandbox needs Docker or NVIDIA OpenShell. Docker covers Python, JavaScript/TypeScript, Go and Ruby, the OpenShell image runs Python and Node, and shell scripts are only analysed statically. Tools that look destructive, such as payments, deletes and network writes, are not run unless you pass allow_write_tools. Each finding includes a fix hint that your coding agent applies itself, so no model or API key is involved.

The README spells out the side effects: a checksum-verified binary download to ~/.trustabl-probe, sandbox image pulls, and four report files written into the scanned directory. It says the server sends no telemetry, makes no model call and never uploads your code. It also never installs Docker, OpenShell or Python itself, and instead returns the command for your agent to ask you about.

Install

Run in a terminal

claude mcp add agent-network-access-probe -- npx -y @trustabl/agent-network-access-probe

Add to claude_desktop_config.json

{
  "mcpServers": {
    "agent-network-access-probe": {
      "command": "npx",
      "args": [
        "-y",
        "@trustabl/agent-network-access-probe"
      ]
    }
  }
}
Install in CursorOpens Cursor, which asks you to confirm. Or add it by hand:

Add to ~/.cursor/mcp.json

{
  "mcpServers": {
    "agent-network-access-probe": {
      "command": "npx",
      "args": [
        "-y",
        "@trustabl/agent-network-access-probe"
      ]
    }
  }
}
Install in VS CodeOpens VS Code, which asks you to confirm. Or add it by hand:

Add to .vscode/mcp.json

{
  "servers": {
    "agent-network-access-probe": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "@trustabl/agent-network-access-probe"
      ]
    }
  }
}

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

{
  "mcpServers": {
    "agent-network-access-probe": {
      "command": "npx",
      "args": [
        "-y",
        "@trustabl/agent-network-access-probe"
      ]
    }
  }
}

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

{
  "mcpServers": {
    "agent-network-access-probe": {
      "command": "npx",
      "args": [
        "-y",
        "@trustabl/agent-network-access-probe"
      ]
    }
  }
}

Tools this server provides (3)

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

probe_check_setup
Reports whether the binary, Docker and sandbox images are ready and gives fix commands.
probe_install
Downloads the verified Probe binary and pulls sandbox images.
probe_scan
Scans a tool directory and returns a score, draft allow list and findings with fix hints.

What it does

Network allow-list drafts
Generates an OpenShell network policy from the tool's code.
Network-blocked sandbox
Runs the tool without network access to expose runtime problems.
Worst-first findings
Each finding comes with a fix hint and a readiness score.
Setup check
Reports whether the binary, Docker and images are ready.
Verified binary download
Checks the binary against release checksums, failing on a mismatch.
Write-tool guard
Skips tools that look destructive unless you allow them.

Good for

  • Draft least-privilege network rules for an agent tool
  • Find missing timeouts or hardcoded credentials in tool code
  • Test a tool for crashes and missing dependencies without network
  • Check sandbox readiness on a new machine

Pricing

Open source

Prices change. Check trustabl.ai before you buy.

Trustabl Probe FAQ

Does it need an API key or model?

No. Your coding agent applies the fixes, and the server makes no model call.

Is my code uploaded?

The README says it is analysed on your machine and never uploaded, and there is no telemetry.

Which languages can the sandbox run?

Docker covers Python, JavaScript/TypeScript, Go and Ruby; OpenShell runs Python and Node.

What does it write to disk?

Four report files in the scanned directory, plus a binary and images installed on request.

Is Trustabl Probe free?

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

How do I add Trustabl Probe to Claude Code?

Run this command in your terminal: claude mcp add agent-network-access-probe -- npx -y @trustabl/agent-network-access-probe

What are the best alternatives to Trustabl Probe?

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

Alternatives to Trustabl Probe

  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