Neo for Claude Code

Use Claude Code as the interface and Neo as the autonomous execution engine for complex AI engineering workflows.

Video walkthrough

Keep one walkthrough video only.

Commands (from Neo MCP flow)

Use the same Claude Code MCP command pattern as the Neo MCP page.

1. Register Neo MCP

claude mcp add --scope user neo \
  -e NEO_SECRET_KEY=sk-v1-YOUR_KEY \
  -- python3 -m neo_mcp

2. Confirm connection

claude mcp list

You should see neo connected in the list.

Claude Code + Neo

Plan deeply, execute iteratively, and ship faster with autonomous workflow control.

Neo is built for end-to-end AI engineering workflows: model building, fine-tuning, evaluation, prompt experiments, and production-oriented ML pipelines. Claude Code stays your interaction layer while Neo handles structured multi-step execution.

Autonomous execution Neo can run multi-step workflows with planning, retries, and iteration loops.Model optionality Switch model strategy by phase as quality and budget constraints change.Engineering focus Build and evaluate AI models, agents, prompts, and ML systems in one flow.

Why use Neo with Claude Code

Keep Claude Code as your primary workspace

Stay in your existing coding flow while delegating heavy AI engineering execution to Neo.

Execute beyond single-pass generation

Neo supports iterative loops for build, evaluate, compare, and improve cycles instead of one-shot outputs.

Optimize quality and cost over time

Use stronger models for strategy and lower-cost models for repetitive execution where appropriate.

Typical workflow

Define goal in Claude Code

Describe the outcome: model benchmark, prompt system, agent, fine-tune job, or pipeline build.

Neo plans execution strategy

Neo expands your objective into an execution plan with concrete steps and validation points.

Neo runs iterative loops

Neo executes, checks outcomes, refines, and reruns where needed to converge on better results.

Review and ship in Claude Code

Inspect outputs, accept changes, and finalize deployment-ready artifacts from your editor flow.

Good fit: projects where quality comes from iteration, not from a single model response.

Use cases this page targets

LLM and prompt evaluation

Compare model responses, track quality tradeoffs, and iterate toward robust prompt systems.

Fine-tuning and model improvement

Run repeatable tuning loops with measurement checkpoints and experiment documentation.

Agent and tool-chain orchestration

Build agents that reason across tools, constraints, and multi-step execution requirements.

Production ML workflows

Develop structured ML pipelines with clearer execution state, iteration paths, and outcomes.

Quick setup

Install and open Claude Code

Start from your normal Claude Code environment and workspace.

Connect Neo workflow access

Configure credentials and execution settings for your Neo environment.

Run a first autonomous task

Use a scoped objective and review how Neo plans and executes in iterative steps.

Tune model strategy by phase

Use higher-capability planning models and efficient execution models as needed.