Machine-readable AI reference
Stable public URLs for AI context, documentation discovery, API schemas, capabilities, and code-level integration.
Files to give your AI
| URL | Format | Purpose |
|---|---|---|
https://docs.torq.finance/ai/torq-context.json | JSON | Compact operation, workflow, source, release, and safety index |
https://docs.torq.finance/ai/torq-agent-system-prompt.txt | Text | Safe default instructions for a Torq-aware assistant |
https://docs.torq.finance/developers.llms.txt | Text | Developer documentation map |
https://docs.torq.finance/llms.txt | Text | Complete public documentation map |
https://docs.torq.finance/openapi/torq-v1.openapi.json | OpenAPI JSON | Exact API methods, auth, request, and response schemas |
https://docs.torq.finance/developer-platform/capabilities.json | JSON | Availability and access class |
These files are public and contain no credentials. Fetch them from docs.torq.finance; do not copy
old versions into a prompt indefinitely.
Human references
- API reference: every governed method and path with exact samples.
- SDK reference: typed clients and signer boundaries.
- Canonical state: freshness and final completion.
- Authentication: publishable, secret, and wallet-scoped access.
- Errors and rate limits: fail-closed handling.
- Status: current Developer Preview availability.
What is not promised
Torq does not currently document a public Torq MCP server, a public autonomous transaction service, or a model that can hold signing authority. Use standard HTTP retrieval and your AI provider's tool interface, with Torq SDK and API calls behind your own allowlisted tools.
Recommended tool allowlist
Give the AI narrow tools for:
- canonical resource reads;
- wallet-scoped reads after session confirmation;
- one typed workflow preparation method;
- workflow and transaction status reads.
Keep key management, raw RPC, arbitrary HTTP, arbitrary contract calls, and signing outside the model toolset.