llms.txt
# IQ Routing
> Vendor-agnostic LLM gateway. Routes each request to a model across your connected providers, caches repeat requests at zero cost, and gives agent loops a cost envelope, an audit trail, and stable capability handles instead of pinned model ids. OpenAI- and Anthropic-compatible at the edge.
Base URL (gateway API): `https://gateway.iq-routing.com`
Dashboard: `https://iq-routing.com`
Auth header: `Authorization: Bearer <key>` -- org API keys are prefixed `gw_live_`
Providers: 19 bring-your-own-key providers, including OpenAI, Anthropic, and Google; see /docs/providers for the full list.
Generated 2026-10-08 from repository documentation sources and approved plan terms.
Free includes IQ Code's cost meter, cache clock, plan pace, budgets, and themes, plus gateway API routing with an IQ API key. IQ Code routing needs Personal ($6/month), with 2,000 routing decisions a month and rough session cost estimates. Your connected providers bill API model usage separately.
For the full condensed reference in one file, see [/llms-full.txt](https://iq-routing.com/llms-full.txt).
## Core concepts
- [Getting started](https://iq-routing.com/docs/getting-started): connect a provider, create an IQ API key, and send a test request.
- [Quickstart](https://iq-routing.com/docs/quickstart): three-line base-URL swap for the OpenAI or Anthropic SDK.
- [Auth](https://iq-routing.com/docs/auth): bearer keys, scopes, rate limits and budgets, revocation.
- [Routing](https://iq-routing.com/docs/routing): routing aliases (including `dynamic`), fallback chain, family lock.
- [Providers](https://iq-routing.com/docs/providers): bring-your-own-key setup, the provider list, custom endpoints.
- [Models](https://iq-routing.com/docs/models): chat model ids by provider.
- [Agent step labels](https://iq-routing.com/docs/step-classifier): Labels for each agent-loop step, so you can see how a loop's work and spend break down.
- [Capability aliases](https://iq-routing.com/docs/capabilities): `cap:` handles that resolve to a concrete model instead of a pinned id.
- [Agent sessions](https://iq-routing.com/docs/sessions): one row per agent run, trajectory replay, cost envelopes.
- [Caching](https://iq-routing.com/docs/caching): the gateway's own response cache versus the provider's prompt cache.
- [Billing](https://iq-routing.com/docs/billing): token classes, thinking-budget tiers, rate-limit scopes, provider-credit exhaustion.
## Tooling
- [IQ Code](https://iq-routing.com/docs/iq-code): Claude Code plugin with free session tools; routing needs Personal.
- [Key vault](https://iq-routing.com/docs/vault): recover plaintext for issued API keys, PATs, and provider credentials.
- [Token optimizer](https://iq-routing.com/docs/optimizer): payload compression before the upstream call, fails open.
- [Webhooks](https://iq-routing.com/docs/webhooks): signed audit-event deliveries to a SIEM, chat, or incident pipeline.
## Security and compliance
- [Security](https://iq-routing.com/docs/security): the authoritative Zero Data Retention boundary, anomaly rate-limits, access scoping.
- [Compliance](https://iq-routing.com/docs/compliance): audit chain, SOC2 evidence pack, Merkle backup, GDPR DSAR, HIPAA posture.
## Operations
- [Operations](https://iq-routing.com/docs/operations): exactly-once scheduled-job guarantees, dependency currency.
- [Webhook capacity planning](https://iq-routing.com/docs/operations/webhook-capacity-planning): sizing a receiver against the delivery envelope.
## SDK recipes
- [SDK index](https://iq-routing.com/docs/sdks): OpenAI, Anthropic, curl, LangChain, Cursor, Codex CLI, and Claude Code.
## Reference
- [API reference](https://iq-routing.com/docs/api-reference): public `/v1` paths, generated from the OpenAPI schema.
- [FAQ](https://iq-routing.com/docs/faq): prompt storage, Zero Data Retention, supported models, pricing.
## Optional
- [Pricing](https://iq-routing.com/pricing)
- [Demo](https://iq-routing.com/demo)
