Neo — AI/ML Engineering
AI/ML engineering MCP server for Neo — submit tasks, track execution, and retrieve outputs.
Connector by NeoAIResearch · GitHub account verified by the MCP Registry
Chalk Labs
Need help deploying Neo — AI/ML Engineering?
Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.
Install
Works with Claude Code, Codex, VS Code, Cursor, Windsurf, Claude Desktop. Read the repository before you give any tool access to your data.
npx agentmarket install connector/neoairesearch-neo-mcp
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this server can do and what it asks for, from its tool list, install config and code. Checked Nov 1, 2026.
Runs code on install
Its package runs an install script (such as npm postinstall or setup.py code) as soon as you install it, before any tool is called. Read the script first.
- Tools
- 8 read · 6 write · 5 delete
- Secrets
- 1 (API keys, tokens or passwords)
- Install scripts
- Runs code on install
- Runs as
- Python package
Tools by what they do 19
Read from the source code; the running server may list different tools.
Read 8 · Look things up without changing anything.
neo_get_execution_evidenceneo_get_messagesneo_list_byok_modelsneo_list_byok_profilesneo_list_integrationsneo_list_tasksneo_task_statusneo_test_integration
Write 6 · Create, change, send or run things.
neo_add_byok_profileneo_add_integrationneo_pause_taskneo_resume_taskneo_send_feedbackneo_set_byok_profile
Delete 5 · Delete, drop, revoke or wipe things.
neo_remove_byok_profileneo_remove_integrationneo_stop_taskneo_submit_taskneo_verify_task
Secrets and settings it asks for
Secrets: values you paste in that let it act as you.
NEO_SECRET_KEY
Other settings (environment variables or headers).
NEO_ENVIRONMENTNEO_WORKSPACE_DIR
Network hosts it talks to 15
alpha.heyneo.com· found in codeapi.anthropic.com· found in codeapi.github.com· found in codeapi.openai.com· found in codeapi.wandb.ai· found in codeapp.heyneo.so· found in codegithub.com· found in codeheyneo.so· found in codehuggingface.co· found in codekaggle.com· found in codemaster.heyneo.com· found in codemaster.heyneo.so· found in codemcpserver.heyneo.com· found in codeopenrouter.ai· found in codes3.amazonaws.com· found in code
Files it touches
~/.git-credentials/private/tmp/proc/meminfo- Your home directory (~)
~/.cache/huggingface/token~/.neo~/.neo/daemon~/.neo/daemon/neo-mcp.log~/.neo/integrations~/.neo/integrations.json~/.neo/settings.jsonC:\\Windows\\System32\\cmd.exeC:\\Windows\\System32\\WindowsPowerShell\\v1.0\\powershell.exe- Any path the agent passes to its tools
Tools (19)
Read from the source code; the running server may differ. Descriptions are the server’s own words; hidden characters are shown as [U+…].
neo_add_byok_profileUSE THIS TOOL when the user wants Neo to run on their own LLM key — phrasings like "use my Anthropic key for Neo's brain", "run Neo on my own OpenAI key", "bring my own key", "BYOK". This is DIFFERENT from neo_add_integration: integrations give a task subprocess credentials (env vars); a BYOK profile changes which LLM powers Neo's orchestrator itself (sent as the x-llm-* headers on task submit + feedback). The key/model is validated against the provider before saving; an invalid key is rejected. By default the new profile becomes active. After it succeeds, relay the 'safety' message verbatim a…
neo_add_integrationUSE THIS TOOL whenever the user wants to give Neo a credential — phrasings like "save my OpenRouter key for Neo", "use this Anthropic key with Neo", "here's my HuggingFace token", "set my GitHub PAT", or any message where the user pastes an API key / token and wants Neo to use it. DO NOT instead suggest the user create a .env file, export an env var, or edit a config file — this tool IS the correct and only supported way to register credentials for Neo. DO NOT ask follow-up questions before calling this tool if the user already provided both the provider and the key — just call it. Key-forma…
neo_get_execution_evidenceread-onlyRead bounded, thread-scoped observations recorded by the local daemon. Evidence contains hashes and exit metadata, never credential values. Provenance is local_daemon_observed; it is not remote attestation.
neo_get_messagesread-onlyRetrieve the full conversation output from a completed Neo task. Only call when neo_task_status returns COMPLETED. For live progress while RUNNING, call neo_task_status again later. Output is capped at ~80,000 characters (~20,000 tokens). If the response is truncated, paginate backwards using the `before` cursor set to the ISO timestamp of the oldest message on the previous page.
neo_list_byok_modelsread-onlyList the model ids a BYOK provider supports. Prefers Neo's own catalog (the authority on what task submit will accept); if that is unavailable, falls back to the provider's live model API (with an api_key) or a curated list. The response 'source' field says which was used. Discovery only when the user did NOT name a model — never pick the first list entry over an explicit user ID.
neo_list_byok_profilesread-onlyList the configured BYOK ('bring your own key') LLM profiles and show which one is active. A BYOK profile makes Neo run its orchestrator on the USER's own LLM key (Anthropic / OpenAI / OpenRouter) instead of Neo's credits. Returns profile id, name, provider, model, a masked key hint, and the active flag. Never returns the raw key.
neo_list_integrationsread-onlyUSE THIS TOOL when the user asks questions like "which keys have I saved for Neo?", "what integrations are set up?", "do I already have an OpenRouter/Anthropic/GitHub/HF key configured?", or before calling neo_add_integration to check if a credential already exists. Returns the provider name, auth method, when it was added, and which credential files are registered. NEVER returns the secret value itself — safe to show the full response to the user. Covers github, huggingface, anthropic, openrouter.
neo_list_tasksread-onlyList all known Neo tasks with their current live status. Use this when returning to a session — e.g. after closing and reopening Claude Code — to see which tasks are still RUNNING, which are COMPLETED, and which need feedback. Returns tasks sorted newest-first. For each task: thread_id, workspace, status, and last-updated timestamp. After getting thread_ids, use neo_task_status or neo_get_messages to drill into a specific task.
neo_pause_taskPause a running Neo task mid-execution. The task freezes at its current step — neo_task_status will then report WAITING_FOR_FEEDBACK. Safe to call on an already-paused task (no-op). After pausing: if you have new instructions or a correction, call neo_send_feedback right away — it resumes the task and delivers the message in one step. Call neo_resume_task instead only if you want it to continue exactly as before, with nothing new to say. To cancel permanently, use neo_stop_task instead.
neo_remove_byok_profilecan change or delete dataDelete a BYOK profile and its stored key. Irreversible — the user must re-add it via neo_add_byok_profile to use it again. If the deleted profile was active, BYOK is cleared (Neo falls back to its own LLM credentials).
neo_remove_integrationcan change or delete dataUSE THIS TOOL when the user says things like "remove my OpenRouter key from Neo", "delete the GitHub integration", "forget my Anthropic key", or "revoke Neo's access to X". Deletes the native credential file(s) and the metadata entry. Irreversible — the user must re-supply the secret via neo_add_integration to use that provider again. If the user wants to REPLACE a key (rotate it), prefer calling neo_add_integration directly — it overwrites the existing entry.
neo_resume_taskResume a paused Neo task from exactly where it stopped, with NO new input — use only when there is nothing to add or correct. Works after neo_task_status reports WAITING_FOR_FEEDBACK following the pause. Has no effect if the task is already running. Only works after neo_pause_task — to start a new task use neo_submit_task. If you have any instructions, correction, or clarification to give, use neo_send_feedback instead — it resumes the task AND delivers your message in one call; calling neo_resume_task would resume it silently and that message would never reach Neo.
neo_send_feedbackReply to Neo, or hand it new instructions before it continues. Call when neo_task_status returns WAITING_FOR_FEEDBACK — Neo is frozen and waiting on you. This is also the status reported after neo_pause_task. Either way, this single call delivers your message AND resumes the task. After sending, call neo_task_status again to confirm. State requirement: feedback is accepted when the task state is WAITING_FOR_FEEDBACK — not while RUNNING. To course-correct a task that is currently RUNNING, call neo_pause_task first, then call neo_send_feedback right away once neo_task_status reports WAITING_FO…
neo_set_byok_profileSelect which BYOK profile is active, or clear it. Pass a profile_id (from neo_list_byok_profiles) to activate that profile, or null to deactivate BYOK so Neo uses its own default LLM credentials again. The active profile's key is attached to every subsequent neo_submit_task and neo_send_feedback.
neo_stop_taskcan change or delete dataPermanently stop and clean up a Neo task. IRREVERSIBLE — execution context is deleted and the task cannot be resumed. Only call when the user explicitly asks to cancel. To pause temporarily (resumable), use neo_pause_task instead.
Show 4 more tools
neo_submit_taskcan change or delete dataSubmit an AI/ML task to Neo for local execution. Use for: training/fine-tuning models, building AI agents, RAG pipelines, LLM integrations, ML data processing. NOT for general coding — write that code directly. Execution is entirely local: the daemon runs on the user's machine and writes files directly to workspace. Files are never stored remotely. Neo's output may reference /app/project/src/model.py — the daemon automatically remaps this to <workspace>/src/model.py on the local disk. Returns {thread_id, status, workspace} immediately. Next: neo_task_status to poll, neo_get_messages when C…
neo_task_statusread-onlyFetch a compact, live backend-reported status for a Neo task. This is remote telemetry, not independently verified execution evidence. Returns one of: RUNNING (still executing — call neo_task_status again later), COMPLETED (done — call neo_get_messages for output), WAITING_FOR_FEEDBACK (Neo is frozen, waiting on you — this is also the status neo_task_status reports after neo_pause_task): call neo_send_feedback if you have new instructions, a correction, or an answer (it resumes and delivers the message in one step), otherwise call neo_resume_task to continue unchanged, TERMINATED or FAILED (en…
neo_test_integrationread-onlyUSE THIS TOOL when the user says "check if my OpenRouter key still works", "test my GitHub token", "is my Anthropic key valid?", or right after neo_add_integration to confirm the credential is live. Also use it FIRST when a Neo task fails with a 401/403/auth error — verify the stored credential before debugging the task. Calls the provider's own API (GitHub, HuggingFace, Anthropic, OpenRouter) and returns {ok, message, latency_ms}. Read-only — does not modify the stored credential.
neo_verify_taskcan change or delete dataIndependently verify local task artifacts using caller-supplied file assertions and optional bounded verification commands. Backend messages alone never produce VERIFIED. Verification commands execute locally and may modify files, so this tool is unavailable in NEO_READ_ONLY mode.
Signals
- Stars
- 2
- Forks
- 1
- Open issues
- 0
- Weekly downloads
- 96
- Last commit (this folder)
- Sep 15, 2026 (1 month ago)
- Last release
- Jun 8, 2026 (4 months ago)
- Last package publish
- Sep 15, 2026 (1 month ago)
From GitHub (fetched Oct 24, 2026) and the package registry.
Details
- Registry name
io.github.NeoAIResearch/neo-mcp- Version
- 0.5.5
- Auth
- API key
- Packages
- pypi neo-mcp
- License
- MIT
- Repository
- github.com/NeoAIResearch/neo-mcp
- Website
- heyneo.com
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
Done for you
Rather not set it up yourself?
Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.
See how we build AI agents →Read the case studies
Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.
- launches shipped
- 50+
- launches shipped
- media placements
- 500+
- media placements
- founders on every account
- 2
- founders on every account
Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block
Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.
Get Neo — AI/ML Engineering running in your business
Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.
Cite this page
Neo — AI/ML Engineering MCP server. Agent Mart. https://agentmart.dev/connectors/neoairesearch-neo-mcp (last updated 2026-10-24)