Back to Browse

Neo MCP Server

Developer ToolsUse Caution4.2MCP RegistryLocal
Free

Server data from the Official MCP Registry

AI/ML engineering MCP server for Neo — submit tasks, track execution, and retrieve outputs.

About

AI/ML engineering MCP server for Neo — submit tasks, track execution, and retrieve outputs.

Security Report

4.2
Use Caution4.2High Risk

Neo MCP is a well-structured tool integration server with proper API key authentication and reasonable permission scoping for its purpose. However, it has moderate security concerns around subprocess command execution (path traversal in command strings, incomplete remapping), insufficient input validation on file paths, and some error handling gaps. The server's ability to execute arbitrary subprocess commands with limited sanitization, combined with path remapping complexities that may bypass isolation, creates exploitable attack vectors. Users should be aware these issues may allow tasks to escape workspace boundaries. Supply chain analysis found 4 known vulnerabilities in dependencies (2 critical, 2 high severity). Package verification found 1 issue.

4 files analyzed · 14 issues found

Security scores are indicators to help you make informed decisions, not guarantees. Always review permissions before connecting any MCP server.

Permissions Required

This plugin requests these system permissions. Most are normal for its category.

File System Read

Reads files on your machine. Normal for tools that analyze or process local data.

File System Write

Writes or modifies files on your machine. Check that this is expected for the tool.

HTTP Network Access

Connects to external APIs or services over the internet.

env_vars

Check that this permission is expected for this type of plugin.

Shell Command Execution

Runs commands on your machine. Be cautious — only use if you trust this plugin.

process_spawn

Check that this permission is expected for this type of plugin.

clipboard

Check that this permission is expected for this type of plugin.

What You'll Need

Set these up before or after installing:

Your Neo API secret key (starts with sk-v1-). Get one at https://heyneo.com.Required

Environment variable: NEO_SECRET_KEY

Neo backend environment: 'production' (default) or 'staging'.Optional

Environment variable: NEO_ENVIRONMENT

Project root where Neo tasks read and write files. Must be the git/project root.Optional

Environment variable: NEO_WORKSPACE_DIR

How to Install

Add this to your MCP configuration file:

{
  "mcpServers": {
    "io-github-neoairesearch-neo-mcp": {
      "env": {
        "NEO_SECRET_KEY": "your-neo-secret-key-here",
        "NEO_ENVIRONMENT": "your-neo-environment-here",
        "NEO_WORKSPACE_DIR": "your-neo-workspace-dir-here"
      },
      "args": [
        "-y",
        "neo-mcp"
      ],
      "command": "npx"
    }
  }
}

Documentation

View on GitHub

From the project's GitHub README.

Neo MCP

Connect your AI editor to Neo's remote execution backend. Describe an AI/ML task in plain language — Neo trains the model, builds the pipeline, or runs the workload on its backend, then writes all output files directly to your local machine.

Works with Claude Code, Cursor, Windsurf, VS Code (GitHub Copilot), Zed, Continue.dev, and OpenAI Codex CLI.

Get your API key at heyneo.com/dashboard → Settings → API Keys.


How it works

Your editor  ──MCP──▶  neo-mcp server  ──API──▶  Neo backend
                            │                          │
                            └──────────────────▶  Local daemon
                                                  (writes files,
                                                   runs scripts)
  1. You describe a task: "Train a fraud detection model on data.csv"
  2. The editor calls neo_submit_task via MCP
  3. Neo's backend processes the task and sends commands to the local daemon
  4. The daemon runs on your machine — writing files, running scripts, installing packages
  5. Output files appear directly in your workspace

Files are always written to your machine, never stored remotely.

Neo can also hold your third-party API keys (GitHub, HuggingFace, Anthropic, OpenRouter, OpenAI, AWS S3, Weights & Biases, Kaggle) locally so it can use them in tasks without re-prompting — keys stay on your machine, never sent to Neo's backend. Full guide: docs/INTEGRATIONS.md.


Install

Pick pip (Python 3.11+) or npm (Node.js 18+) — both work identically from the editor's perspective.

pip

pip install neo-mcp

Use pipx install neo-mcp to avoid virtualenv conflicts.

npm

npm install -g neo-mcp

Install Node.js 18+ if needed: curl -fsSL https://deb.nodesource.com/setup_20.x | bash - && apt-get install -y nodejs


Connect to your editor

Replace sk-v1-YOUR_KEY with your actual API key. Pick the pip or npm command based on which you installed.

Claude Code

pip:

claude mcp add --scope user neo \
  -e NEO_SECRET_KEY=sk-v1-YOUR_KEY \
  -- neo-mcp

npm:

claude mcp add --scope user neo \
  -e NEO_SECRET_KEY=sk-v1-YOUR_KEY \
  -- neo-mcp-daemon --mcp

Open a new Claude Code session after running — tools load at session start, not mid-session.


Cursor

Open the config:

  • GUI: Ctrl+Shift+J (Windows/Linux) or Cmd+Shift+J (Mac) → Tools & MCP → New MCP Server
  • Or edit the file directly: ~/.cursor/mcp.json

pip:

{
  "mcpServers": {
    "neo": {
      "command": "neo-mcp",
      "env": { "NEO_SECRET_KEY": "sk-v1-YOUR_KEY" }
    }
  }
}

npm:

{
  "mcpServers": {
    "neo": {
      "command": "neo-mcp-daemon",
      "args": ["--mcp"],
      "env": { "NEO_SECRET_KEY": "sk-v1-YOUR_KEY" }
    }
  }
}

Restart Cursor after editing the file directly. Changes via the GUI apply immediately.


Windsurf

Open the config:

  • GUI: Ctrl+, (Windows/Linux) or Cmd+, (Mac) → Cascade → Plugins (MCP servers) → Manage Plugins → View raw config
  • Or edit the file directly: ~/.codeium/windsurf/mcp_config.json

pip:

{
  "mcpServers": {
    "neo": {
      "command": "neo-mcp",
      "env": { "NEO_SECRET_KEY": "sk-v1-YOUR_KEY" }
    }
  }
}

npm:

{
  "mcpServers": {
    "neo": {
      "command": "neo-mcp-daemon",
      "args": ["--mcp"],
      "env": { "NEO_SECRET_KEY": "sk-v1-YOUR_KEY" }
    }
  }
}

Changes apply on save — no restart needed.


VS Code (GitHub Copilot)

Requires VS Code 1.99+.

Open the config:

  • Command Palette: Ctrl+Shift+P (Windows/Linux) or Cmd+Shift+P (Mac) → "Chat: Open Chat Customizations"
  • Or edit the file directly: .vscode/mcp.json in your workspace root (create if it doesn't exist)

pip:

{
  "servers": {
    "neo": {
      "type": "stdio",
      "command": "neo-mcp",
      "env": { "NEO_SECRET_KEY": "sk-v1-YOUR_KEY" }
    }
  }
}

npm:

{
  "servers": {
    "neo": {
      "type": "stdio",
      "command": "neo-mcp-daemon",
      "args": ["--mcp"],
      "env": { "NEO_SECRET_KEY": "sk-v1-YOUR_KEY" }
    }
  }
}

MCP tools only work in Copilot Agent mode — switch to Agent mode in the chat panel.


Zed

Open the config:

  • Ctrl+Alt+, (Windows/Linux) or Cmd+Alt+, (Mac) — opens settings.json directly
  • Or: Command Palette → "zed: open settings"

Add inside ~/.config/zed/settings.json:

pip:

{
  "context_servers": {
    "neo": {
      "source": "custom",
      "command": {
        "path": "neo-mcp",
        "args": [],
        "env": { "NEO_SECRET_KEY": "sk-v1-YOUR_KEY" }
      }
    }
  }
}

npm:

{
  "context_servers": {
    "neo": {
      "source": "custom",
      "command": {
        "path": "neo-mcp-daemon",
        "args": ["--mcp"],
        "env": { "NEO_SECRET_KEY": "sk-v1-YOUR_KEY" }
      }
    }
  }
}

Changes apply on save — no restart needed.


Continue.dev

Open the config:

  • Ctrl+L (VS Code) or Ctrl+J (JetBrains) to open the sidebar → click Agent selector above the chat input → click the gear icon
  • Or edit the file directly: ~/.continue/config.yaml

pip:

mcpServers:
  - name: neo
    command: neo-mcp
    env:
      NEO_SECRET_KEY: sk-v1-YOUR_KEY

npm:

mcpServers:
  - name: neo
    command: neo-mcp-daemon
    args:
      - --mcp
    env:
      NEO_SECRET_KEY: sk-v1-YOUR_KEY

MCP tools only work in Agent mode — switch to Agent in the mode selector.


OpenAI Codex CLI

Open the config:

  • Run codex mcp to manage servers interactively via CLI
  • Or edit the file directly: ~/.codex/config.json

pip:

{
  "mcpServers": {
    "neo": {
      "command": "neo-mcp",
      "env": { "NEO_SECRET_KEY": "sk-v1-YOUR_KEY" }
    }
  }
}

npm:

{
  "mcpServers": {
    "neo": {
      "command": "neo-mcp-daemon",
      "args": ["--mcp"],
      "env": { "NEO_SECRET_KEY": "sk-v1-YOUR_KEY" }
    }
  }
}

Verify the connection

Claude Code:

claude mcp list

You should see neo with a green checkmark. Then open a new session and ask:

"What Neo tools do you have available?"

The assistant should list neo_submit_task, neo_task_status, neo_get_messages, and more.


Tools

ToolDescription
neo_submit_taskSubmit a task to Neo. Returns thread_id immediately.
neo_list_tasksList all running and recent tasks — useful after reopening your editor.
neo_task_statusCheck status: RUNNING / COMPLETED / WAITING_FOR_FEEDBACK / PAUSED / TERMINATED.
neo_get_messagesRead full task output once status is COMPLETED. Capped at ~20 000 tokens.
neo_send_feedbackReply to Neo when it asks a clarifying question (WAITING_FOR_FEEDBACK).
neo_pause_taskPause a running task. Can be resumed.
neo_resume_taskResume a paused task.
neo_stop_taskPermanently stop and clean up a task.
neo_list_integrationsList stored third-party API keys (names only — never the value).
neo_add_integrationRegister a credential (GitHub, HuggingFace, Anthropic, OpenRouter, OpenAI, AWS S3, Weights & Biases, Kaggle) so Neo tasks can use it as an env var.
neo_test_integrationCall the provider's API to confirm a stored key is still valid.
neo_remove_integrationDelete a stored key from this machine.

Integration tools store credentials locally — file mode 0o600 under ~/.neo/integrations/ (or native tool files like ~/.aws/credentials, ~/.netrc, ~/.kaggle/kaggle.json), or your OS keyring if NEO_INTEGRATIONS_BACKEND=keyring. Keys never leave your machine. Full guide: docs/INTEGRATIONS.md.


Workflow

Standard (tasks over a few minutes):

neo_submit_task  →  returns thread_id
      ↓
neo_task_status  →  poll until COMPLETED or WAITING_FOR_FEEDBACK
      ↓
neo_get_messages →  read full output

Quick task: Pass wait_for_completion: true to neo_submit_task — it blocks until done and returns output directly. No polling needed.

Mid-task question: When status is WAITING_FOR_FEEDBACK, call neo_send_feedback with your reply. Neo resumes automatically.

Reconnecting after closing your editor:

neo_list_tasks   →  all tasks with live status + thread IDs
neo_task_status  →  check the specific task you care about
neo_get_messages →  read output of any COMPLETED task

Diagnostics (pip)

neo-mcp status      # daemon and key status
neo-mcp doctor      # full health check — identifies common issues
neo-mcp list        # list known threads
neo-mcp logs --source neo-mcp --lines 100   # MCP server logs
neo-mcp logs --source daemon --lines 100    # daemon logs

# JSON output
neo-mcp status --json
neo-mcp doctor --json

Claude Code logs:

claude mcp logs neo

Environment variables

VariableRequiredDescription
NEO_SECRET_KEYYesAPI key (sk-v1-...) from heyneo.com/dashboard → Settings → API Keys
NEO_DEPLOYMENT_IDNoPin a specific daemon UUID (auto-generated by default)
NEO_WORKSPACE_DIRNoOverride workspace directory (useful in Docker or CI)
NEO_READ_ONLYNotrue = expose only status/message tools — disables submit, stop, and pause

Troubleshooting

SymptomFix
neo-mcp: command not foundRe-run pip install neo-mcp, verify with which neo-mcp
neo-mcp-daemon: command not foundRe-run npm install -g neo-mcp, verify with which neo-mcp-daemon
✗ Failed to connect in claude mcp listRun claude mcp logs neo — most common cause is NEO_SECRET_KEY not set
Neo tools don't appearOpen a new session — tools load at session start, not mid-session
Invalid API key (401)Re-check your key at heyneo.com/dashboard → Settings → API Keys
Trial or quota ended (403)Top up at the Neo dashboard
No healthy deployments available (400)Daemon failed to auto-start — restart the MCP server and try again
Task submitted but no files writtenDaemon stopped mid-task — check neo-mcp status and restart
Status stuck on RUNNINGRun neo-mcp doctor to diagnose; restart the MCP server
Output truncated~20 000 token cap — use neo_task_status for progress, neo_get_messages for final output only

Example prompts

Train a fraud detection model on fraud.csv, optimize for recall
Build a sentiment analysis pipeline for product reviews and save the model
Analyse sales_data.csv and produce a feature importance report
Fine-tune a text classifier on my training data with 5-fold cross-validation

Neo handles the ML execution — your editor handles everything else.


For a full setup guide including all editor options: docs/GUIDE.md

Reviews

No reviews yet

Be the first to review this server!