API

The Spore API is OpenAI-compatible. Create an API key from your profile’s API Keys tab, then point your existing OpenAI client at Spore.

Base URL

https://api.sporeintel.com/api/v1

Authentication

Authorization: Bearer sk_spore_YOUR_API_KEY

List models

curl https://api.sporeintel.com/api/v1/models \
  -H "Authorization: Bearer sk_spore_YOUR_API_KEY"

Model names include a context tier, e.g. qwen3.5-4b:8k or qwen3.6-35b-a3b:32k. Append :personal to any model name to route the request to your own nodes only. Personal requests are free.

Each entry includes recommended, catalog_role (when set), and available. Recommended models are listed first. Previous model ids remain valid and stay listed; available is false when the model cannot currently accept a request.

Chat completions

curl https://api.sporeintel.com/api/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer sk_spore_YOUR_API_KEY" \
  -d '{
    "model": "qwen3.6-35b-a3b:32k",
    "messages": [
      {"role": "user", "content": "Hello!"}
    ],
    "stream": false
  }'

Set "stream": true for server-sent events; stream_options: {"include_usage": true} adds token usage to the final chunk, matching the OpenAI streaming format.

To keep a completion so you can fetch it later, add "store": true — see Retrieving completions.

Reasoning

Many models can reason (“think”) before answering. Control it with the OpenAI-standard reasoning_effort:

curl https://api.sporeintel.com/api/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer sk_spore_YOUR_API_KEY" \
  -d '{
    "model": "qwen3.6-27b:32k",
    "messages": [{"role": "user", "content": "How many rs are in strawberry?"}],
    "reasoning_effort": "low"
  }'

• reasoning_effort accepts none, minimal, low, medium, or high. none turns reasoning off on models that support toggling it; any other value turns it on. Models with graded effort (e.g. gpt-oss) use the level; always-on models ignore none.
• The reasoning text is returned separately in message.reasoning_content (and delta.reasoning_content when streaming), so content stays clean.
• A boolean enable_thinking is also accepted and takes precedence over reasoning_effort when both are sent.

Reasoning support is per model. Each entry from /api/v1/models includes supports_thinking, thinking_default_on, and thinking_toggleable so you can detect it programmatically.

Retrieving completions

By default, completions aren’t retained — the response is returned once and then discarded. Send "store": true to keep a completion so you can fetch it later, and optionally attach metadata (up to 16 key–value pairs) for filtering:

curl https://api.sporeintel.com/api/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer sk_spore_YOUR_API_KEY" \
  -d '{
    "model": "qwen3.5-4b:8k",
    "messages": [{"role": "user", "content": "Hello!"}],
    "store": true,
    "metadata": {"project": "demo"}
  }'

Stored completions are kept for your account’s data-retention window (30 days by default; Pro accounts can configure it). After that the record remains but its content is cleared (content_purged: true on retrieval).

Retrieve one using the id from the completion response:

curl https://api.sporeintel.com/api/v1/chat/completions/CHAT_COMPLETION_ID \
  -H "Authorization: Bearer sk_spore_YOUR_API_KEY"

List your stored completions (newest first). Query params: limit (1–100, default 20), order (asc or desc), model, metadata[KEY]=VALUE, and after (a completion id, for pagination):

curl "https://api.sporeintel.com/api/v1/chat/completions?limit=20&metadata[project]=demo" \
  -H "Authorization: Bearer sk_spore_YOUR_API_KEY"

The response is a list: {"object": "list", "data": [...], "first_id": "...", "last_id": "...", "has_more": false}.

List the input messages of a stored completion:

curl https://api.sporeintel.com/api/v1/chat/completions/CHAT_COMPLETION_ID/messages \
  -H "Authorization: Bearer sk_spore_YOUR_API_KEY"

Update a stored completion’s metadata (the only mutable field):

curl https://api.sporeintel.com/api/v1/chat/completions/CHAT_COMPLETION_ID \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer sk_spore_YOUR_API_KEY" \
  -d '{"metadata": {"project": "archived"}}'

Delete a stored completion (removes it and clears its content):

curl -X DELETE https://api.sporeintel.com/api/v1/chat/completions/CHAT_COMPLETION_ID \
  -H "Authorization: Bearer sk_spore_YOUR_API_KEY"

Requests you submit with Spore’s end-to-end encryption are stored as ciphertext; retrieval returns that ciphertext with is_encrypted: true, since the server can’t decrypt it.

Using coding tools

Any tool that speaks the OpenAI protocol works: set its base URL to https://api.sporeintel.com/api/v1 and use your sk_spore_ key. The app’s Code page generates ready-to-paste configuration (including an OpenCode config with per-model context limits) from the models currently on the network.

Cursor desktop

Cursor can use Spore through its OpenAI-compatible override. Create a sk_spore_ key from the API Keys tab (shown once at creation; mint a new key if you no longer have it), then:

1. Open Cursor Settings and go to Models.

2. Enable OpenAI API Key and paste your sk_spore_ key. The field is labeled OpenAI even for compatible providers.

3. Enable Override OpenAI Base URL and paste:

   https://api.sporeintel.com/api/v1

4. Click + Add Model and type a Spore model id verbatim (name:tier from GET /models). Append :personal to any model ID to route directly to your own node. No credits charged.

5. Click Verify. Cursor hits https://api.sporeintel.com/api/v1/models to check the connection.

Tab completion and Cmd-K stay on Cursor’s backend. Only chat that uses this OpenAI-compatible override is sent to Spore.

The stock OpenAI-compatible path is plaintext. It is not the community end-to-end encrypted chat path.

Override OpenAI Base URL is global. It hijacks Cursor’s Pro and built-in models until you turn it off. Leave the override off when using Cursor’s own models.

The in-app Code page walks through the same setup with the model ids currently on the network.

Pricing

Requests are charged in credits by model, context tier, and token counts. See Credits & Pricing. Personal-mode requests are free.