Probe MCP server
Commit Nov 14, 2026Allows coding agents to read machine learning experiment runs, metrics, and lineage to avoid repeating previous team efforts.
Also known as prbe-ai/probe
- Stars
- 0
- Tools
- —
- Downloads, 30 days
- —
- Latest version
- 1.0.1
- Last commit
- Nov 14, 2026
- License
- Apache-2.0
What is Probe MCP?
Probe allows coding agents to read machine learning experiment runs, metrics, and lineage to avoid repeating previous team efforts.
Summary generated from the project's README
About Probe MCP server
Probe provides machine learning experiment tracking designed for both human teams and coding agents. It allows AI assistants to access experiment runs, metrics, and lineage data so they do not repeat work already completed by others. This tool helps agents understand how specific results were achieved and enables them to continue experiments exactly where a previous user or agent left off.
Developers can integrate this server into various coding agents including Claude Code, Codex, Cursor, pi, and Kimi Code. Installation is handled through the coding agent or by running npx probe-research. For Python-based training scripts, the probe-research SDK is available via pip. Users can also use the instrument-code skill to wire the SDK directly into their training code using npx skills add prbe-ai/probe.
What you can do with it
- 1Prevent redundant experimentationAn AI agent can check previous experiment results to ensure it does not attempt a training run that has already failed or succeeded.
- 2Analyze experiment lineageA developer can ask an agent to explain the specific parameters and steps used to produce a particular machine learning result.
- 3Hand off active experimentsA user can instruct a coding agent to take over a machine learning task by resuming from the last recorded state.
Written from the project's README. Every use case is checked against a quote from it.
How to install Probe MCP
Formats verified 10 Nov 2026Add to CursorAdd to GooseInstall Probe MCP in Claude Code
$ claude mcp add --transport http probe https://mcp.research.prbe.ai/mcpInstall Probe MCP in Cursor
~/.cursor/mcp.json
{
"mcpServers": {
"probe": {
"url": "https://mcp.research.prbe.ai/mcp"
}
}
}Install Probe MCP in VS Code
.vscode/mcp.json
{
"servers": {
"probe": {
"type": "http",
"url": "https://mcp.research.prbe.ai/mcp"
}
}
}Install Probe MCP in VS Code CLI
$ code --add-mcp '{"name":"probe","type":"http","url":"https://mcp.research.prbe.ai/mcp"}'Install Probe MCP in Codex
$ codex mcp add probe --url https://mcp.research.prbe.ai/mcpInstall Probe MCP in Codex config.toml
~/.codex/config.toml
[mcp_servers.probe] url = "https://mcp.research.prbe.ai/mcp"
Install Probe MCP in Gemini CLI
$ gemini mcp add -s user --transport http probe https://mcp.research.prbe.ai/mcpInstall Probe MCP in Gemini settings.json
~/.gemini/settings.json
{
"mcpServers": {
"probe": {
"url": "https://mcp.research.prbe.ai/mcp",
"type": "http"
}
}
}Automated checks
Ran Nov 15, 2026Full report →Not verified yet. A server needs a score of 80, no serious finding, and enough data for us to check. How checks work · What to fix for more points
- Source repository linked and reachableprbe-ai/probe is public and reachablegithub.com/prbe-ai/probe ↗Pass
- Package names this repository as its sourceNot checked
- Build provenance (Sigstore attestation)Not checked
- Open source license (OSI)Apache-2.0spdx.org/licenses/Apache-2.0.html ↗Pass
- Commit or release in the last yearLast commit or release 2026-11-14github.com/prbe-ai/probe ↗Pass
- OpenSSF ScorecardNot checked
- Package name not close to a popular packageNot checked
- No critical advisory in the package or its dependenciesNot checked
- No hidden instructions to the model0 tool descriptions and the README read; no override phrases, credential paths or hidden charactersPass
- Destructive tools are labelledNot checked
- Remote endpoint answers the MCP handshakeHTTPS endpoint answered 401 (sign-in required)Pass
README
View on GitHubProbe
Experiment tracking for ML teams and the coding agents that run their experiments: Claude Code, Codex, Cursor, pi and Kimi Code. Keep working the way you do. From now on:
- You'll always know how any result was made.
- Your agent won't repeat what your team already tried.
- You can pick up any experiment where it left off, or hand it off.
Set it up from your coding agent. Paste this:
Install Probe on this machine and its coding agents. Read https://research.prbe.ai/install.md and follow it.
Or run npx probe-research yourself. For a training script, pip install probe-research gives
you the Python SDK. To add only the skill that wires the SDK into training code:
npx skills add prbe-ai/probe --skill instrument-code.
This repository
This repository is a CI-generated distribution mirror. Nobody commits here: every commit is pushed by the release pipeline of Probe's private monorepo, and the entire tree is derivable from that repo at any commit. Pull requests and hand-pushed commits are not accepted — the next sync would overwrite them.
What it serves:
- Claude Code / Codex plugin marketplace —
probe@probe(experiment-tracking skills + read-only MCP wiring),probe-daemon@probe(the daemon profile: the Probe daemon records and reads, installed byprobe wizardin place ofprobe@probe) andprobe-tap@probe(client-sanitized transcript capture). Installed plugins update from this repo'smain. pi installs theplugins/probe-research-pipackage from this repo's rootpackage.json, and Kimi Code installs these plugins from a checkout of it. skills/instrument-code— the skill that wires the Probe SDK into training code, on its own, fornpx skills add(rendered from theprobe@probeplugin).client-version.json— the version/staleness advisory manifest the Probe backend reads over GitHub raw.CHANGELOG.mdand the Releases page — the public release record for theprobe-researchCLI/SDK (PyPI,npx probe-researchon npm).
Issues and contributions: the source of truth is private; report problems via your Probe contact or the support channels on the site above. Licensed under Apache-2.0 (see LICENSE).
Probe MCP questions
How do I add Probe MCP to Claude Code?
Run claude mcp add --transport http probe https://mcp.research.prbe.ai/mcp in your terminal. Add --scope user to make it available in every project.
How do I add Probe MCP to Cursor?
Open ~/.cursor/mcp.json and add the server under mcpServers, using the Cursor config shown in the install section above.
What can coding agents do with Probe?
Coding agents can read your team's ML runs, metrics, and lineage to understand how results were achieved and avoid repeating previous experiments.
How do I install Probe for a training script?
You can install the Python SDK using pip install probe-research.