Smart Spawn — OpenClaw Plugin

Intelligent model routing for OpenClaw. Automatically picks the best AI model for any task based on real benchmark data from 5 sources.

Instead of hardcoding models or guessing, Smart Spawn analyzes what you're doing and routes to the optimal model — factoring in task type, budget, benchmarks, speed, and your own feedback history.

Install

openclaw plugins install @deeflectcom/smart-spawn
openclaw gateway restart

That's it. The plugin talks to the public API at ss.deeflect.com — no self-hosting needed.

Usage

Just use your agent normally. Smart Spawn adds a smart_spawn tool that the agent calls when spawning sub-agents:

"Research the latest developments in WebGPU"

→ Picks Gemini 2.5 Flash (fast, free, great context) and spawns a research sub-agent

"Build me a React dashboard with auth"

→ Picks the best coding model in your budget tier and spawns a coder sub-agent

Spawn Modes

• Single — Pick the best model, spawn one agent
• Collective — Pick N diverse models, run in parallel, merge results
• Cascade — Start cheap, escalate to premium if needed
• Plan — Decompose sequential multi-step tasks and assign best model per step
• Swarm — Decompose complex tasks into a DAG of sub-tasks with optimal model per step

Configuration

Optional — add to your OpenClaw config under plugins.entries.smart-spawn.config:

{
  "apiUrl": "https://ss.deeflect.com/api",
  "defaultBudget": "medium",
  "defaultMode": "single"
}

| Setting | Default | Options | |---------|---------|---------| | apiUrl | https://ss.deeflect.com/api | Your own API URL if self-hosting | | defaultBudget | medium | low, medium, high, any | | defaultMode | single | single, collective, cascade, plan, swarm | | collectiveCount | 3 | Number of models for collective mode (2-5) | | telemetryOptIn | false | Opt-in to anonymous community telemetry | | communityUrl | apiUrl | Community API URL for shared telemetry |

How It Works

The plugin calls the Model Intelligence API which:

1. Pulls model data from 5 sources — OpenRouter, Artificial Analysis, HuggingFace Leaderboard, LMArena (Chatbot Arena), LiveBench
2. Z-score normalizes benchmarks across sources so scores are comparable
3. Scores models per category — coding, reasoning, creative, vision, research, fast-cheap, general
4. Factors in budget tiers, speed, cost efficiency, and your personal feedback
5. Returns the optimal model for the task

Budget Tiers

| Budget | Price Range (per 1M input tokens) | |--------|----------------------------------| | low | $0 – $1 | | medium | $0 – $5 | | high | $2 – $20 | | any | No limit |

Self-Hosting the API

The API source is in the same repo. If you want full control, run your own instance and point apiUrl to it.

License

MIT — @deeflect