Neo MCP — AI engineering, without leaving your editor

Stay in Claude Code, Cursor, or Codex and hand your AI/ML work to Neo — an autonomous AI engineer that plans, builds, runs, and evaluates entire workflows from a plain-English request. neo-mcp is the Model Context Protocol server that connects Neo to the AI coding tools you already use, so you never switch tabs: ask your agent to send a task to Neo, and a local daemon writes the resulting code, models, metrics, and reports straight into your repo — on your machine, nothing stored remotely.

Because Neo is purpose-optimized for AI engineering — not a general-purpose coding assistant — it goes deeper on ML, LLM, and data workflows than a general coding agent can.

📚 Docs: https://docs.heyneo.com/neo-mcp · 🔑 Get an API key: Neo dashboard

What MCP unlocks

🧩 Stay in your editor — drive Neo from Claude Code, Cursor, VS Code (Copilot), Windsurf, Zed, Continue, or Codex. No new app, no context switching.
🔬 More depth — autonomous planning, experiments, evaluation, and iteration tuned for AI/ML, beyond what a generic coding agent attempts.
💾 Local-first — every output file is written to your machine; nothing is stored remotely.

What you can build with Neo

🤖 Generative AI & LLMs — RAG & semantic search, agents & chatbots, fine-tuning (Llama, Qwen, Gemma), document analysis
🧠 ML & deep learning — PyTorch / TensorFlow / scikit-learn training, architecture search, evaluation
📊 Data science & analytics — EDA, feature engineering, forecasting, segmentation, A/B testing, reporting
👁️ Computer vision — image classification, object detection, OCR
🎤 Speech & audio — speech-to-text, text-to-speech, audio classification
🔌 Bring your own keys — GitHub, HuggingFace, Anthropic, OpenRouter, OpenAI, AWS S3, Weights & Biases, Kaggle — stored locally, injected as env vars

For data scientists, ML & LLM engineers, analysts, researchers, and PMs who want results, not boilerplate.

Install

pip install neo-mcp

Requires Python 3.11+ on Linux, macOS, or Windows through WSL. Native Windows cmd.exe/PowerShell execution is intentionally unsupported because Neo commands use POSIX shell and container-path semantics.

Tip: use pipx install neo-mcp to install in an isolated environment and avoid conflicts with your project's virtualenv.

Use Neo from your editor

Neo runs in every major MCP-enabled AI editor — set it up once below. Replace sk-v1-YOUR_KEY with your actual key.

Claude Code

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

Open a new Claude Code session after running this. Neo tools load at session start.

Scope options: --scope user (global, recommended) · --scope project (writes .mcp.json in current repo) · --scope local (this machine only)

Verify it registered:

claude mcp list

Cursor

~/.cursor/mcp.json:

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

Restart Cursor after editing.

Windsurf

~/.codeium/windsurf/mcp_config.json:

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

Restart Windsurf after editing.

VS Code (GitHub Copilot)

.vscode/mcp.json in your workspace root (requires VS Code 1.99+):

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

Zed

~/.config/zed/settings.json:

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

Continue.dev

~/.continue/config.json:

{
  "mcpServers": [
    {
      "name": "neo",
      "transport": {
        "type": "stdio",
        "command": "neo-mcp",
        "env": {
          "NEO_SECRET_KEY": "sk-v1-YOUR_KEY"
        }
      }
    }
  ]
}

OpenAI Codex CLI

~/.codex/config.json:

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

Tools

Tool	Description
`neo_submit_task`	Submit a locally destructive AI/ML task; one active owner per workspace.
`neo_list_tasks`	List running and recent tasks — reconnects pollers automatically.
`neo_task_status`	Compact live backend-reported status; telemetry is not independent proof.
`neo_get_messages`	Read full task output when COMPLETED. Capped at ~20 000 tokens.
`neo_send_feedback`	Reply when Neo asks a question (WAITING_FOR_FEEDBACK).
`neo_pause_task`	Pause a running task.
`neo_resume_task`	Resume a paused task.
`neo_stop_task`	Stop and clean up a task permanently.
`neo_list_integrations`	List stored third-party API keys (names only — never the value).
`neo_add_integration`	Register a GitHub PAT / HuggingFace token / Anthropic key / OpenRouter key so Neo tasks can use it as an env var.
`neo_test_integration`	Call the provider's API to confirm a stored key is still valid.
`neo_remove_integration`	Delete a stored key from this machine.
`neo_get_execution_evidence`	Read bounded, hash-chained local execution observations.
`neo_verify_task`	Verify local artifacts with explicit file assertions and acceptance commands.

Integration tools store credentials locally (file 0o600 under ~/.neo/integrations/, or OS keyring with NEO_INTEGRATIONS_BACKEND=keyring). Keys never leave your machine. See the full guide at docs/INTEGRATIONS.md.

Environment variables

Variable	Required	Description
`NEO_SECRET_KEY`	Yes	Your Neo API key (`sk-v1-...`) from the Neo dashboard
`NEO_DEPLOYMENT_ID`	No	Pin a specific deployment UUID (auto-generated and persisted by default)
`NEO_WORKSPACE_DIR`	No	Override working directory (useful in Docker)
`NEO_READ_ONLY`	No	`true` — expose only status, messages, task/evidence reads, integration-test/list, and BYOK-list tools; reject every mutation and verification command
`NEO_INIT_CHAT_TIMEOUT_SECONDS`	No	Submit timeout in seconds; defaults to `60` to match the production VS Code client
`NEO_POLLER_READY_TIMEOUT_SECONDS`	No	Maximum pre-submit wait for the local daemon to establish its backend poll connection; defaults to `15`

Troubleshooting

Symptom	Fix
`neo-mcp: command not found`	Re-run `pip install neo-mcp` and verify your PATH with `which neo-mcp`
Tools don't appear after registering	Open a new session — MCP tools load at session start, not mid-session
`Invalid API key` (401)	Re-check your key in the Neo dashboard
`Trial or quota ended` (403)	Top up at the Neo dashboard
Task submitted but no files written	Daemon failed to start — run `neo-mcp doctor` to diagnose
Status stuck on RUNNING	Call `neo_task_status` to check; run `neo-mcp status` to inspect the daemon

Release files for neo-mcp 0.5.14

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for neo-mcp 0.5.14
File	Size	Uploaded
neo_mcp-0.5.14.tar.gz	177.5 kB	Sep 7, 2026	Details

Built distribution (wheel)

Table of built distributions (wheels) for neo-mcp 0.5.14
File	Interpreter	ABI	Platform	Reset
neo_mcp-0.5.14-py3-none-any.whl 138.5 kB Sep 7, 2026	Python 3	none	any	Details

Total release size: 316.0 kB

Release files / neo_mcp-0.5.14.tar.gz

Download URL	neo_mcp-0.5.14.tar.gz
Size	177.5 kB
Tags	Source
SHA-256 checksum How to use checksums	`6cf2138a6908c83f38b6b9bcdceaad771911b9468332b1ce8a1a4a6e1afc0ccc`
BLAKE2b-256 checksum How to use checksums	`01e1f5ba34fd44783454002c4ce116d46fca79bb3ff0d440c4d56e5cb197643b`
Upload date	Sep 7, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	No
Uploaded via	`twine/7.0.0 CPython/3.13.9`

Release files / neo_mcp-0.5.14-py3-none-any.whl

Download URL	neo_mcp-0.5.14-py3-none-any.whl
Size	138.5 kB
Tags	Python 3
SHA-256 checksum How to use checksums	`e3b1b7c980451268928bac309b8d0ff7e1038e7454f3daccb9dbac068c367f73`
BLAKE2b-256 checksum How to use checksums	`40d144ada6c9cd624ca1537913b303fc007cccadc8ac48c7d2b54d931d09c450`
Upload date	Sep 7, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	No
Uploaded via	`twine/7.0.0 CPython/3.13.9`