Agent integrations
Use an LLMBase Pro or Business chat subscription from OpenAI-compatible agents such as OpenClaw and Hermes.
Updated
LLMBase Pro subscribers and active Business seats can use their chat subscription from OpenAI-compatible agents. Free and Starter plans cannot create or use chat agent keys.
Use this integration when you want OpenClaw, Hermes, or another OpenAI-compatible agent to consume your LLMBase Pro or Business chat subscription. Use the direct Inference API instead when you are building a product, backend service, or high-volume automation that needs prepaid token-based billing.
Key docs
| Page | Use it for |
|---|---|
| Models for agents | Model availability, capabilities, and recommended starting models |
| Usage and rate limits | Included Chat/Agent budget, request estimates, request limits, and production workload guidance |
| Setup examples | Creating a chat agent key, configuring custom OpenAI-compatible clients, Hermes, OpenAI SDK, and cURL |
| Discovery metadata | Well-known JSON entrypoints that help agents discover public LLMBase capabilities |
Authentication model
You sign in to the LLMBase dashboard with your LLMBase account, then create a dedicated chat agent key for external agents. OpenClaw, Hermes, and other OpenAI-compatible tools do not sign in with your LLMBase account directly. They authenticate by sending the chat agent key as a Bearer token.
The OAuth and OIDC endpoints documented in Discovery metadata
are discovery metadata for agents and compatible clients. Current chat-agent
access does not use an OAuth, browser-login, or device-code flow. If a tool asks
for a custom OpenAI-compatible provider, use the agent API base URL and your
llmbase_chat_... key.
Chat agent keys vs inference API keys
Chat agent keys are different from inference API keys:
| Key prefix | Use this for | Base URL | Billing source |
|---|---|---|---|
llmbase_chat_... | OpenClaw, Hermes, and other external agents | https://llmbase.ai/api/v1/agents | Included Pro or Business Chat/Agent subscription budget |
llmbase_... | Direct inference API integrations | https://api.llmbase.ai/v1 | Prepaid credits |
Agent requests use the same included subscription budget as normal LLMBase Chat. Inference API usage is billed separately with prepaid credits.
- Revoking one key does not revoke the other product’s keys.
- Chat-agent keys do not have per-key spend caps. They are subscription access credentials; the included Chat/Agent budget applies at the account level.
- Inference API keys cannot access subscription-only chat-agent endpoints.
Agent API
Chat agent keys work against the agent API base URL:
- Base URL:
https://llmbase.ai/api/v1/agents - Models:
GET /models - Chat completions:
POST /chat/completions - Auth:
Authorization: Bearer <LLMBASE_CHAT_AGENT_KEY>
The older https://llmbase.ai/api/chat-agent/v1 base URL remains available
for existing clients, but new integrations should use /api/v1/agents.
LLMBase keeps this endpoint separate from the direct inference endpoint on
purpose. https://api.llmbase.ai/v1 is for API integrations with llmbase_...
inference keys, using prepaid credits.
https://llmbase.ai/api/v1/agents is for OpenClaw, Hermes, and other external
agents that should consume a user’s Pro or Business chat subscription with a
llmbase_chat_... key.
If the account is not on Pro and does not have an active Business seat, the API
returns 403 with chat_pro_required. Upgrade to Pro or join a Business plan
before creating or using chat agent keys.
Only inference models are accessible via agents; browse the current list in inference models or read Models for agents.
What is included in Pro and Business
| Use case | Included in Pro or Business chat agent access? | Notes |
|---|---|---|
| OpenClaw or Hermes configured as a custom OpenAI-compatible provider | Yes | Use https://llmbase.ai/api/v1/agents and a llmbase_chat_... key |
| Interactive coding agents, research assistants, and self-hosted assistants | Yes | Uses the same included subscription budget as browser chat |
| Long-running autonomous loops | Limited | Supported technically, but bounded by included budget and server-side rate limits |
| API-like backend workloads for your own app or customers | No | Use the direct Inference API with a llmbase_... key |
| Separate token-billed usage accounting | No | Use the direct Inference API with a llmbase_... key |
Agent access is intended for personal or team assistant workflows: coding agents, local developer tools, research helpers, and self-hosted assistants that send user-initiated tasks. It is not intended to replace the prepaid Inference API for unattended high-volume services, resale, customer-facing API products, or continuously running automation that behaves like infrastructure traffic.
Read Usage and rate limits before using agents for long loops or repeated background work.
Privacy and EU compliance
LLMBase is built for European users and GDPR-conscious workflows. Chat-agent access uses the same account and included-budget contract as LLMBase Chat. It does not spend prepaid Inference API credits.
For compliance-sensitive workflows:
- review and accept the LLMBase DPA from Dashboard -> Legal & Compliance when needed
- avoid sending secrets or regulated personal data unless your legal basis and data-processing setup are in place
- choose models and workflows that match your data-residency requirements
- use the direct Inference API for production systems that need explicit token-level accounting and operational separation from your chat subscription
This documentation is technical product guidance, not legal advice.
FAQ
Can I sign in to OpenClaw with my LLMBase account?
No. LLMBase account login is used in the dashboard to manage your subscription and keys. OpenClaw connects to LLMBase through the OpenAI-compatible API using a chat agent key.
Why can OpenClaw sign in with an OpenAI or ChatGPT account?
That is a special OpenAI Codex integration in OpenClaw. It is separate from the generic OpenAI-compatible connection flow. For custom endpoints such as LLMBase, OpenClaw uses a base URL plus API key.
Should I use my inference API key with agents?
No. Use a llmbase_chat_... chat agent key for OpenClaw, Hermes, and other
chat-agent integrations. Use a llmbase_... inference API key only for the
LLMBase inference API.
Can I set a spending limit per chat-agent key?
No. Chat-agent keys are part of the Pro and Business chat subscription products,
so quota is account-level instead of key-level. You can revoke individual
chat-agent keys, and usage draws from the shared included Chat/Agent subscription budget. Use a
llmbase_... inference API key when you need per-key monthly spend caps.
Will LLMBase add OAuth or account sign-in for agents?
The current supported integration is Bearer-key authentication because it works with the widest range of OpenAI-compatible tools. OAuth or device login may be added later as a convenience layer, but chat agent keys remain the compatible path for external agents today.