Model groups for agents.
A single key provides access for OpenClaw, Hermes, Pi, Conductor, and other agents to GLM, MiniMax, DeepSeek, Claude, Codex, GPT-5.6, and Fable (with usage limits determined by plan).
Open, agent, and premium model groups from €39/mo.
GLM, MiniMax, and DeepSeek for the quick passes. Claude or Codex when the step is heavier. Same endpoint and dashboard either way.
Effort Levels
Effort (low, medium, high) is available right away. Set it per request depending on whether you want speed or more care.
Your Own Dashboard
Analytics that update live, a request feed, breakdowns by model, and token counts. See input, output, cache reads, and writes for each call.
Works With Everything
One key works with OpenClaw, Hermes, Pi, Conductor, OpenCode, Claude Code CLI, Codex Desktop, Cursor, direct API, and the rest that accept a custom base URL.
No More 4-Hour Cooldowns
People on single accounts sometimes get cut off for hours after a short burst. OpenLimits spreads the load so the session keeps going even on heavy days.
Built for developers who ship
Production routing infrastructure with detailed observability.
Full Streaming
Server-sent event streaming with no additional wrapper layer. Tokens appear as soon as they are generated by the provider.
Token Analytics
Input, output, cache reads, and writes counted per request. Trends, estimates, and model splits live in the dashboard.
Model Breakdown
Usage split by model with daily trends and savings estimates. Charts and tables update as you switch between models.
Live Feed
See requests arrive as they happen. Filter by model and look at any single call for model, tokens, latency, and status.
Zero Downtime
Requests are not subject to cooldown windows or forced waits. When one account reaches a limit, traffic is automatically routed to the next available account.
Native Agent Model Access
GLM, MiniMax, DeepSeek, Claude, Codex, GPT-5.6, and Fable show up through one key and one model list.
Extended Thinking
Extended thinking works with a budget you set. Claude can step through hard problems and you see the process while it streams.
PDF & Image Support
Send PDFs, images, and other files in the same requests you would send to the direct API.
Works with every tool you use
Set one environment variable. Every tool picks it up automatically.
Conductor
It picks up your Claude Code CLI env vars. Or add ANTHROPIC_BASE_URL and ANTHROPIC_API_KEY in Settings → Env.
Claude Code CLI
Two lines in your shell profile and every session goes through OpenLimits. Extended thinking, streaming, and tool use work the same.
OpenCode
It reads the same environment variables on start. No extra files or setting changes needed.
Cursor
Add OpenLimits in the provider settings. Use the Claude models without the usual subscription caps.
OpenClaw
Point it at the base URL with your key. The rest of OpenClaw works as before.
Direct API & SDKs
The Anthropic and OpenAI SDKs, cURL, and any client that takes a custom base URL. Examples are in the API docs.
OpenLimits vs Claude Max + Codex Pro
Access to both model families without the per-service caps, at a lower combined price.
| Claude Max ($100-200/mo) | Codex Pro ($200/mo) | OpenLimits (€39/mo) | |
|---|---|---|---|
| Claude models | Yes (with limits) | No | Yes, unlimited |
| Codex/GPT models | No | Yes (with limits) | Yes, unlimited |
| Rate limits | 5x/20x Pro limits | Per-model caps | None |
| Cooldowns | Still hits caps | Still hits caps | Never |
| Effort levels | No | No | Yes |
| Extended thinking | Limited | Limited | Unlimited |
| API access | No | Yes | Full REST API |
| Token analytics | No | No | Full dashboard |
| Works with dev tools | Web only | Codex Desktop | CLI, Cursor, Codex, SDKs, etc. |
| Automatic failover | N/A | N/A | Yes |
| Combined price | $300-400/mo | €39/mo | |
How it is built
The parts you might care about when something goes wrong.
Edge Deployed
Persistent Node.js services on dedicated VPS infrastructure keep streaming connections stable for long-running responses.
Legitimate & Secure
Enterprise accounts are purchased directly from the providers. No borrowed keys are used. All traffic is encrypted in transit. Prompts and completions are not stored. API keys are scoped to individual accounts.
Real-Time Tracking
Model, token counts, latency, and status are recorded for every call. The live feed uses server-sent events.
99.9% Uptime
Edge nodes plus automatic failover. When the providers are up, we stay up.
Stop hitting limits. Start shipping.
Starting at €39/month. GLM, MiniMax, DeepSeek, Claude, Codex, GPT-5.6, and Fable where the plan allows. Set up in 30 seconds.
Get Your API Key →