# Autonomous Relay Agents > Bring your own provider keys to one OpenAI-compatible endpoint for every model. Automatic failover, per-key spend limits, PII redaction. 0% markup. Autonomous Relay Agents (https://autonomousrelay.com) is an OpenAI-compatible API gateway. One base URL, https://autonomousrelay.com/v1, routes chat completions to 294+ models from 16+ providers, served on the caller's own provider keys at the provider's list price. Any OpenAI SDK works by changing the base URL and sending an Autonomous Relay Agents inference key as a Bearer token. Streaming, failover between providers, spend ceilings and guardrails (PII redaction, content filters) are part of the same endpoint. Autonomous Relay Agents is made and operated by Tech Service 4U Inc., a Canadian company. ## OpenAI-Compatible LLM API Documentation - [OpenAI-Compatible LLM API Documentation](https://autonomousrelay.com/docs): One /v1 endpoint for every provider. Point your OpenAI client at a new base URL and key, bring your own provider keys, and pay 0% markup on tokens. - [Quickstart](https://autonomousrelay.com/docs/quickstart): Your first request in five steps, then the OpenAI SDK for Python and Node.js and curl, pointed at one endpoint for Claude, Gemini, GPT and every other model. - [Authentication](https://autonomousrelay.com/docs/authentication): The two Autonomous Relay Agents key classes, inference and management: how each authenticates, how keys are handled and managed, and how to check one. - [Chat completions](https://autonomousrelay.com/docs/chat): The chat completions endpoint: parameters, model fallback, the response shape, app attribution, and how to inspect a request after it ran. - [Streaming](https://autonomousrelay.com/docs/streaming): Stream chat completions over server-sent events with Autonomous Relay Agents, and the three behaviours that most often catch integrators out. - [Routing](https://autonomousrelay.com/docs/routing): How Autonomous Relay Agents routes a request: the provider fields, preferences that reorder endpoints, constraints that reject them, and guardrails that outrank both. - [Models & providers](https://autonomousrelay.com/docs/models): Read an Autonomous Relay Agents catalogue entry: context length, per-token prices, cache prices, supported parameters and the provider that serves each endpoint. - [Bring your own key](https://autonomousrelay.com/docs/byok): Bring your own provider keys to Autonomous Relay Agents: what happens to a key, adding one, the fallback policy when it fails, and managing credentials. - [What we store](https://autonomousrelay.com/docs/data): What Autonomous Relay stores and for how long: no prompts or responses by default, request metadata for 24 months, and how to delete what you stored. - [Errors](https://autonomousrelay.com/docs/errors): Every status code the Autonomous Relay Agents API returns, what each one means for your request, and which of them are worth retrying. - [Limits & credits](https://autonomousrelay.com/docs/limits): Autonomous Relay Agents rate limits, spend ceilings, client workspaces and reports, member roles, agent budgets, usage alerts, credits and usage reporting. - [Spend limits and the ledger](https://autonomousrelay.com/docs/spend-limits): Per-key spending limits for LLM APIs on your own provider keys: key, agent, workspace and guardrail budgets, hard or soft, and a ledger you can audit. - [LLM guardrails: PII redaction, allowlists and regions](https://autonomousrelay.com/docs/guardrails): Guardrails for LLM traffic: redact PII, secrets and prompt injections before a prompt leaves, and restrict models, providers and data regions. - [LLM fallback and provider failover](https://autonomousrelay.com/docs/failover): LLM fallback chains and provider failover: the models array, account-wide fallbacks, which errors and timeouts move on, and why spend limits still hold. - [Server tools and MCP servers](https://autonomousrelay.com/docs/tools): Tools the gateway runs inside a request: the built-in server tools, your own MCP servers, marketplace plugins, and how the tool loop, streaming and limits behave. - [Structured outputs](https://autonomousrelay.com/docs/structured-outputs): Ask any model for JSON with response_format: what OpenAI, Gemini and Anthropic each receive, routing to endpoints that support it, and the response-healing plugin. - [Batch API](https://autonomousrelay.com/docs/batches): Submit up to 10,000 chat completions to a provider's own batch API at its batch price on your own key: the request shape, polling, results, cancelling and billing. - [Response cache](https://autonomousrelay.com/docs/caching): The exact-match response cache: turning it on per key, what counts as a hit, the x-cache header, what is never cached, limits, and emptying it. - [Observability export](https://autonomousrelay.com/docs/observability): Send a metadata record of every request to a signed webhook, an OpenTelemetry collector, S3-compatible storage, Langfuse or Datadog: the record, the headers, the API. - [Billing & the ledger](https://autonomousrelay.com/docs/billing): How each request is charged: metered lines with snapshotted prices, your own provider keys, zero-output requests, holds, and CSV exports. - [API reference](https://autonomousrelay.com/docs/reference): Every Autonomous Relay Agents endpoint with its authentication, parameters, request body and responses, generated from the published OpenAPI description. - [OpenAI SDK](https://autonomousrelay.com/docs/openai-sdk): Use OpenAI's Python and Node SDKs with any catalogue model: the base URL, environment variables, which methods work, fallbacks through extra_body, and retries. - [relay launcher](https://autonomousrelay.com/docs/relay-cli): Run Claude Code on your own provider keys with relay: browser sign-in, the key it receives, the variables it sets, and its commands. - [Cursor](https://autonomousrelay.com/docs/cursor): Use any catalogue model in Cursor chat and agent on your own provider keys: the OpenAI key, the base URL override and custom model ids, step by step. - [Cline](https://autonomousrelay.com/docs/cline): Point Cline at Autonomous Relay Agents with its OpenAI Compatible provider: base URL, inference key, model id, and the model limits to set by hand. - [Continue](https://autonomousrelay.com/docs/continue): Add any catalogue model to Continue in config.yaml: the openai provider with apiBase, a key from a .env file, and tool use for agent mode. - [LangChain](https://autonomousrelay.com/docs/langchain): Use Autonomous Relay Agents from LangChain in Python and JavaScript: ChatOpenAI with a base URL, an inference key and a catalogue model id. - [Vercel AI SDK](https://autonomousrelay.com/docs/vercel-ai-sdk): Use Autonomous Relay Agents with the Vercel AI SDK: createOpenAICompatible with our base URL, then generateText and streamText on any model. - [n8n](https://autonomousrelay.com/docs/n8n): Use any catalogue model in n8n workflows and AI agents: an OpenAI credential with our base URL, and the chat model node with the Responses API off. - [Single sign-on (SSO)](https://autonomousrelay.com/docs/sso): Set up single sign-on with Okta, Microsoft Entra ID or Google Workspace, step by step: the OIDC app, a DNS-verified domain, a test sign-in and activation. - [Directory sync (SCIM)](https://autonomousrelay.com/docs/scim): Provision and deprovision people from Okta or Microsoft Entra ID with SCIM 2.0, step by step, and map directory groups to client workspaces. - [Workload identity federation](https://autonomousrelay.com/docs/federation): Call the API from GitHub Actions, Google Cloud or Kubernetes without a stored API key: exchange the workload's identity token for a short-lived token, step by step. - [Building plugins](https://autonomousrelay.com/docs/plugins): Build a plugin for the marketplace: host an MCP server, choose API-key or OAuth sign-in, write agent tools, test it privately and submit it for review. - [Self-hosted deployment](https://autonomousrelay.com/docs/self-host): Run Autonomous Relay Agents in your own network under an Enterprise agreement: Docker Compose or Helm, your Postgres and KMS key, no telemetry. - [Security & disclosure](https://autonomousrelay.com/docs/security): Report a security issue in Autonomous Relay Agents: the contact address, the PGP key fingerprint, what is in scope, testing rules and the safe harbour. ## Product - [Autonomous Relay Agents](https://autonomousrelay.com/): Bring your own provider keys to one OpenAI-compatible endpoint for every model. Automatic failover, per-key spend limits, PII redaction. 0% markup. - [LLM API Pricing Comparison](https://autonomousrelay.com/models): Compare LLMs from every provider by price per million tokens, context window and modality, and call any of them through one OpenAI-compatible API at 0% markup. - [LLM API Providers: HQ, Privacy & Terms](https://autonomousrelay.com/providers): Compare the LLM API providers behind one OpenAI-compatible endpoint: HQ country, privacy policy, terms and status page for each. Bring your own keys. - [Model rankings](https://autonomousrelay.com/rankings): The models Autonomous Relay Agents customers use most, by tokens processed today, this week and this month, and which are rising. Use, not quality. - [Compare models](https://autonomousrelay.com/compare): Put up to four models side by side: prices, context and output limits, modalities, tool and structured-output support, who serves them, and measured uptime. - [Model collections](https://autonomousrelay.com/collections): The catalogue by what models do: free models, million-token context, vision, documents, audio, image and video generation, speech, embeddings and rerankers. - [LLM Gateway Pricing: Free BYOK, 0% Markup](https://autonomousrelay.com/pricing): Solo is free; Team is $99 a month per organisation after a 14-day trial. Inference runs on your own provider keys at list price: 0% markup, no per-request fee. - [Enterprise](https://autonomousrelay.com/enterprise): Single sign-on, a private routing tier, extended ledger retention and a service-level agreement — agreed in a contract rather than toggled in a dashboard. - [relay: Claude Code on Your Own Keys](https://autonomousrelay.com/launcher): relay starts Claude Code on Autonomous Relay Agents: browser sign-in, your own provider keys, your guardrails and budgets, every request tagged in Activity. - [Developers: SDKs, Agents, MCP and the API](https://autonomousrelay.com/developers): Every way into Autonomous Relay Agents: the OpenAI SDK, Vercel AI SDK, LangChain, n8n, Cursor, Cline, Continue, the relay launcher, MCP, and the OpenAPI and ARD documents. - [LLM Playground: Try Any Model](https://autonomousrelay.com/chat): Send a request through Autonomous Relay Agents from the browser and watch which provider serves it, what it cost and how long it took. - [Changelog](https://autonomousrelay.com/changelog): Dated changes to the Autonomous Relay Agents API and model catalogue. New endpoints, routing changes and catalogue updates that affect how you integrate. - [Status](https://autonomousrelay.com/status): Whether the Autonomous Relay Agents API is answering right now, and where to check the status page each upstream provider publishes for itself. ## Optional - [Full documentation](https://autonomousrelay.com/llms-full.txt): every documentation page above in full, as one text file. - [About Tech Service 4U Inc.](https://autonomousrelay.com/about): Who operates Autonomous Relay Agents: Tech Service 4U Inc., a Manitoba corporation, and what the gateway does and does not do with your requests. - [Support](https://autonomousrelay.com/support): How to reach Autonomous Relay Agents support: the status page and error reference first, then support@autonomousrelay.com, with what to include and where security and privacy requests go. - [Privacy Policy](https://autonomousrelay.com/privacy): What we collect, why, who else sees it, and what you can make us do about it. - [Terms of Service](https://autonomousrelay.com/terms): The rules for using Autonomous Relay Agents: what you may build, what you may not, how billing works, and who is responsible for what. - [Sitemap](https://autonomousrelay.com/sitemap.xml): every page above, and one page per model in the catalogue.