GOAL: API key in under 5 minutes · AUDIENCE: developers · CTA: Get API key (primary), Open playground, Read the docs. Code first, prose second.

CAIRN API · v2026-09

Your first call in five minutes.

One endpoint, three models, citations as structured data. OpenAPI spec, typed SDKs and an MCP server included.

Get API key Open playground

$5 free credit · no card for the first key · Read the docs →

~/demo · zsh

$ curl https://api.cairn.example/v1/responses -H "Authorization: Bearer $CAIRN_API_KEY" -d '{"model":"ridge-2.0","input":"Which contracts renew before Q4?"}'

{
  "model": "ridge-2.0",
  "output_text": "Three vendor contracts auto-renew before October 1…",
  "citations": [{ "n": 1, "source": "vendor-register.xlsx" }],
  "verification": "grounded",
  "usage": { "input_tokens": 1204, "output_tokens": 318 },
  "latency_ms": 812
}

Quickstart

STEP 1 · 30 s

Install the SDK

pip install cairn

npm install @cairn/sdk

# no install needed

STEP 2 · 1 min

Set your API key

export CAIRN_API_KEY="ck_live_…"

export CAIRN_API_KEY="ck_live_…"

export CAIRN_API_KEY="ck_live_…"

STEP 3 · 2 min

Make your first call

from cairn import Cairn

client = Cairn()
r = client.responses.create(
    model="ridge-2.0",
    input="Hello, Cairn",
)
print(r.output_text, r.citations)

import Cairn from "@cairn/sdk";

const client = new Cairn();
const r = await client.responses.create({
  model: "ridge-2.0",
  input: "Hello, Cairn",
});
console.log(r.outputText, r.citations);

curl https://api.cairn.example/v1/responses \
  -H "Authorization: Bearer $CAIRN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"ridge-2.0",
       "input":"Hello, Cairn"}'

MICRO-CONVERSION: playground run without a key. After the run: “Get the code for this request” converts to signup.

Playground

DEMO · CANNED RESPONSES

Model Temperature0.3

max_tokens 1024 sources ["demo://legal-ops"]

INPUT

⌘↵ to run

◇ OUTPUT · ridge-2.0 READY

Press Run to see a streamed response. No key needed in the playground.

"citations": [
  {"n": 1, "source": "demo://legal-ops/helio-msa.pdf", "span": [212, 318]},
  {"n": 2, "source": "demo://legal-ops/vendor-register.xlsx", "row": 14}
]

Pick by quality, speed and cost

Quality: Cairn GroundedQA v3, n = 2,400, Aug 2026, 95% CI shown. Latency: p95, 1k in / 300 out, eu-central. Illustrative.

Summit 2.1

summit-2.1

Most capable. Long contracts, multi-step reasoning, agents.

Quality (GroundedQA)91.4% ± 1.1

Speed · p951.9 s

Cost · $ / 1M input$12

Read the model card →

Ridge 2.0

ridge-2.0

Balanced. The default for assistants and RAG.

Quality (GroundedQA)87.2% ± 1.3

Speed · p95820 ms

Cost · $ / 1M input$3

Docs →

Pebble 1.4

pebble-1.4

Fast and low cost. Classification, extraction, routing.

Quality (GroundedQA)78.9% ± 1.6

Speed · p95240 ms

Cost · $ / 1M input$0.25

Docs →

SDKs

Rate limits

Tier	Requests / min	Tokens / min	Unlocks
Free	20	40k	First key
Tier 1	500	400k	$5 paid
Tier 2	3,000	2M	$100 paid + 7 days
Enterprise	Custom	Custom	Contract

429 responses include retry-after. Prices: token pricing and cost estimator →

Changelog

Sep 18, 2026

NEW
Citations with character spans

Responses now return exact spans for every citation, so you can highlight the source text.
Sep 02, 2026

IMPROVED
Ridge 2.0 is 18% faster at p95

No change to quality on GroundedQA v3 (87.2% vs 87.0%).
Aug 12, 2026

DEPRECATION
pebble-1.2 retires Dec 1, 2026

Move to pebble-1.4. Same price, same limits. Migration guide in the docs.

Status

APIOperational
AssistantOperational
AgentsOperational
ConsoleOperational

Incident history →