OpenClawDocs
On this page

On this page

View as MarkdownView this page as plain textOpen in ChatGPTAsk questions about this pageOpen in ClaudeAsk questions about this page

Providers

Particle

Particle serves Particle-1 through an OpenAI-compatible Chat Completions API.

Property Value
Provider particle
Plugin @openclaw/particle-provider
API key environment variable PARTICLE_API_KEY
Base URL https://api.particle.ai/v1
API openai-completions
Default model particle/particle-ai/particle-1

Setup

  1. Create an API key at platform.particle.ai. Inference requires the key's inference scope and a positive credit balance unless Particle grants the key free access.

  2. Install the provider plugin:

    bash
    openclaw plugins install @openclaw/particle-provider

  3. Run onboarding:

    bash
    openclaw onboard --auth-choice particle-api-key

For non-interactive setup with a key already in the environment:

bash
openclaw onboard --non-interactive --accept-risk --skip-health \  --auth-choice particle-api-key \  --particle-api-key "$PARTICLE_API_KEY"

Make the environment variable available to the Gateway process, not just your interactive shell. See Environment variables.

Verify configuration and select the model:

bash
openclaw models list --provider particleopenclaw models set particle/particle-ai/particle-1

Particle-1

The catalog uses the ID Particle lists and returns, particle-ai/particle-1. The short alias particle-1 is accepted wherever you type a model.

  • Text and image input; text output, reasoning, and function tool calling.
  • 1,048,576-token context window; up to 131,072 output tokens. Particle defaults to 32,000 output tokens when a client does not send a limit; OpenClaw sends its own output limit.
  • Standard prices: $0.25 input, $0.03 cached input, and $1.00 output per million tokens. Reasoning tokens count as output tokens. These are catalog estimates, not a live billing report.

Limits and prices were checked against Particle's model catalog and billing documentation on October 29, 2026.

Reasoning and caching

Particle-1 defaults to high reasoning. OpenClaw exposes off, minimal, low, medium, high, xhigh, and max:

  • off sends reasoning_effort: "none".
  • minimal and low select low effort.
  • medium, high, and xhigh select high effort.
  • max selects max effort.

See Thinking levels and Particle's reasoning controls.

The plugin enables the standard conversation prompt_cache_key for cache routing. Particle does not document a client-selectable cache-retention duration.

Supported API surface

This integration uses HTTP Chat Completions with SSE streaming. After you sign in, the model picker lists every chat model your Particle account lists (for example deepseek-ai/deepseek-v41-flash-1017 and z-ai/glm-5.3-flash); the curated limits, prices and thinking controls on this page are for Particle-1. It does not integrate Particle's separate Responses WebSocket API, mid-turn steering, file uploads or hosted web search.

Particle accepts JSON mode, but does not advertise schema-guaranteed structured output. Do not rely on strict JSON Schema enforcement.

Official references: Quickstart, Authentication, Chat Completions.

Was this useful?