One API forevery modelyou'll ever ship with.
One API forevery modelyou'll ever ship with.
One OpenAI-compatible endpoint for open-weight, open-source, and frontier models — with adaptive routing, hosted agents, pipelines, and observability built in. Spin up in minutes; we handle the rest.
No credit card · Free tier with 50 requests & 200K tokens a day
POST /v1/chat/completions200 OK
{ "model": "auto", "messages": [{ "role": "user", "content": "Summarise this PR" }], "stream": true }
Routed to
DeepSeek R1
cache · fallbacks · budgets
Built to carry every kind of intelligence
Large language modelsEmbeddings & retrievalVision · audio · imageHosted agentsPipelinesOpenAI-compatible HTTPLarge language modelsEmbeddings & retrievalVision · audio · imageHosted agentsPipelinesOpenAI-compatible HTTP
Models fromDeepSeekMetaAlibabaMistral AINVIDIABAAI
The pillars
Six quiet ways we do the work so your team stays heads-down.
Three steps to production
Less to build. More to ship.
Drop-in compatible with the OpenAI SDK. From signup to your first production call in minutes — keep the client you already use.
01
Authenticate
Create an account, mint a key, set a budget cap.
02
Point your SDK
Drop our base URL into any OpenAI-compatible client.
03
Ship
Route, cache, observe — models hot-swap without redeploys.
cURLPythonNode OpenAI-compatible
from openai import OpenAI
client = OpenAI( base_url="https://api.septorlabs.com/v1", api_key=os.environ["SEPTOR_KEY"], )
resp = client.chat.completions.create( model="auto", # gateway picks the model messages=[{"role": "user", "content": "Hello"}], )
Model catalog
A hundred-odd models.
One reading list.Frontier reasoning, open weights, embeddings, speech, and vision — and the agents that put them to work. All behind one familiar `/v1` endpoint.
01 · ReasoningDeepSeek V4 ProDeepSeek AIReasoningTools164K
02 · FrontierQwen 3.6 MaxAlibabaChatTools256K
03 · FrontierMiniMax M2.5MiniMaxReasoningTools200K
04 · Open weightsKimi K2.6Moonshot AIChatTools256K
05 · Open weightsGLM-5Zhipu AIChatTools128K
06 · FastGemma 4 27BGoogleChatFast128K
07 · VisionQwen 3 VLAlibabaVisionChat128K
08 · ImageFLUX.2Black Forest LabsImage—
09 · EmbeddingsBGE-M3BAAIEmbeddings8K
10 · SpeechParakeet TDT 0.6BNVIDIAAudio30s/seg
A representative selection
Agents
The agent that never sleeps.
Spawn an agent that browses the web or runs a long task in the background. Hosted on our nodes, priced by the hour, with explicit CPU and memory caps and a fail-closed credit gate.
Browser automation
OpenClaw
A headless browser agent that runs sandboxed Chromium and reasons over the rendered DOM — browse, fill forms, scrape behind auth, and capture screenshots. Billed by the hour, not by the seat.
Long-running tasks
Hermes
A persistent async agent: state lives on a mounted volume and a webhook endpoint streams events in, so it survives restarts and runs for hours. Billed by the hour.
Pipelines
Threads of intelligence, stitched together.
Compose models, tools, and your own data into reproducible graphs. Run on demand, on a schedule, or as the response to a webhook — same key, same bill.
I
Retrieval & answer
Ground every response in your knowledge — cited, dated, and re-runnable.
EmbedSearchAnswer
II
Compare across vendors
Ask two or three models the same question, judge with a third, return the winner.
RouteScorePick
III
Audio in, audio out
Calls, lectures, and voice notes turned into briefs and read back on the commute.
TranscribeSummariseSpeak
Observability
You shipped it. Now watch it run.
Every request through the gateway leaves a structured trace. These are what you get on day one — no instrumentation, no third-party agent to install.
Latency percentiles
P50 and P99 per model, per upstream, and per pipeline node — updated with each request.
Cache analytics
Semantic cache hit rate and savings, shown in real time as your traffic flows through.
Cost attribution
Token cost broken down by model, agent, and pipeline stage. No invoice surprises.
Error visibility
Aggregate and per-call error rates, with structured logs linked from each trace.
Request throughput
Requests per minute and token volume charted across your key, org, and per-model.
Streaming traces
Time-to-first-token and full stream duration recorded for every streaming call.
All metrics reflect your own traffic — no synthetic data shown.
What you can hold us to
No theatre. Just the contract.
We'd rather show you how the product behaves than show you a wall of stock quotes. Here's what holds on every plan.
01
No synthetic data, ever
Every chart reflects your own traffic. We never pad dashboards with invented numbers to make the product look busier.
02
Predictable pricing
Per-token rates published in your dashboard, pre-paid credits, and prices in USD. No surprise invoices at month-end.
03
Guardrails on every tier
Per-key budgets, IP allow-lists, and a full audit log — from the free tier up to enterprise. Spend can't run away from you.
04
Transparent hourly billing
Agents run on managed nodes billed by the hour, with explicit CPU and memory caps and a fail-closed credit gate. You see the rate before you deploy, and you stay in control of when they run.
Simple, honest pricing
Pay only for what you ship.
Open-source models, pre-paid credits, and per-token rates published in your dashboard. No surprise invoices.
MonthlyYearly · save 10%
Free
$0forever
$5 in free credits on signup — make your first production calls, no card required.
- $5 free credits on signup
- Free models: 50 requests & 200K tokens / day
- 1 pipeline
- Open-source catalog Start free
- 10 RPM (burst 1.5×)
- $20 in credits / month
- 2 pipelines
- Standard OSS models Choose Starter
- 20 RPM
- $50 in credits / month
- 4 pipelines
- All OSS models, incl. frontier Choose Paid
- 30 RPM
- $100 in credits / month
- 6 pipelines
- Free OpenClaw agent deploy Choose Pro Plus
- 60 RPM
- $500 in credits / month
- Unlimited pipelines
- Dedicated provider pool
Starter
$20/ mo
For side-projects starting to find a rhythm.
Most popular
Paid
$50/ mo
The everyday plan for teams shipping to users.
Best value
Pro Plus
$100/ mo
More headroom, plus a free agent deploy.
Enterprise
$500/ mo
Dedicated capacity and priority routing.
PAYG credits valid 12 months · Plan credits valid 1 month (no rollover) · Per-key budgets, IP allow-lists, and an audit log on every tier. Full pricing & per-model rates →
Get started today
Start building today.
Start with $5 in free credits on signup — no card required. After that, the free models keep going at 50 requests and 200K tokens a day, with one pipeline and the open-source catalog. Upgrade when you're ready to scale.
By signing up you agree to our Terms and Privacy Policy.
Septor Labs — One API for Every AI Model