P
PulseLABS dev
Pulse Gateway · OpenAI & Anthropic compatible · Every call attested

Change one line.
Every AI call becomes provable.

Point the OpenAI or Anthropic SDK you already use — or Claude Code — at the gateway. PII is masked before egress, the right model is picked per call, and every completion ships a signed certificate a third party can verify offline.

const client = new OpenAI({
-  apiKey: process.env.OPENAI_API_KEY,
+  baseURL: 'https://api.pulse-labs.fr/v1', apiKey: process.env.PULSE_API_KEY,
})
// response headers now include:
x-pulse-proof: pf_a059713a…  // signed, chained, offline-verifiable
x-pulse-pii-masked: 2
Drop in — 5 minutesGet API KeyExplore the engines →
Scroll to explore
Without Synapse Engine
✕Hardcode gpt-4o everywhere, pay full price for every call
✕Personal data flows straight to a third-party model
✕Build your own retrieval pipeline for every document set
✕Write the function-calling loop and signature verification yourself
✕No idea what any single feature actually costs you
✕Re-architect your AI layer every time you add a provider
VS
Synapse Engine
✓
Routes every call to the cheapest model that can do the job
task: 'simple' | 'complex' | 'auto'
✓
Redacts emails, IBANs, phone numbers before they leave your infra
sensitive: true · auto-escalates on detection
✓
Retrieves your own documents and injects them into context
POST /v1/knowledge-sources
✓
Lets agents call your endpoints, HMAC-signed like a webhook
POST /v1/tools
✓
Logs the provider, model, and savings of every single request
GET /synapse/analytics
✓
Same code on OpenAI, Anthropic, Groq, DeepSeek, or local Ollama
provider: openai | anthropic | groq | ollama

"If you only need a single model behind a single API key, don't buy us. If you need routing, redaction, retrieval, and tool execution to just work, no matter which provider is behind it — that's what Synapse Engine is for."

One request, end to end
1
Request
task · source
2
PII Shield
sensitive: auto-escalates
3
Model Router
cheapest capable model wins
4
Knowledge + Tools
POST /v1/knowledge-sources · POST /v1/tools
5
Restore + Log
SynapseLog · GET /synapse/analytics
Step 1 of 5

Request

Your call arrives with a task type — simple, complex, extraction, code, or auto — and an optional knowledge or tools flag. No model name. That's the point.

The Engine
Debate Engine
POST /v1/conversations
SimEngine
POST /v1/simulations
Risk Matrix
POST /v1/risk-sessions
Market Analysis
POST /v1/market-analysis
Agent Personas
POST /v1/agents
Knowledge Base (RAG)
POST /v1/knowledge-sources
Tools (Function Calling)
POST /v1/tools

Debate Engine

STRATEGY · CONSENSUS

Spin up multi-agent debate rounds where AI personas argue, challenge, and synthesize on any topic. Returns consensus scores, devil's advocate positions, and structured synthesis you can pipe directly into your product.

import { PulseLabsSDK } from '@pulselabs/sdk'

const pulse = new PulseLabsSDK({ apiKey: process.env.PULSELABS_API_KEY! })

const { conversation } = await pulse.conversations.create('Market entry: APAC vs LATAM')

const result = await pulse.conversations.sendMessage(
  conversation.id,
  'Should we enter APAC or LATAM first?',
)
// → { synthesis: { consensus: 0.72, recommendation: '...' }, rounds: [...] }
Sample Output
Strategist
APAC has lower CAC and higher NPS benchmark.
Skeptic
Regulatory overhead adds 4–6 months minimum.
Synthesis
Consensus 0.72 — Stage APAC Q3, use LATAM as hedge.
Build in minutes

One key. Every decision module.

curl -X POST https://api.pulse-labs.fr/v1/conversations \
  -H "x-api-key: sk_live_your_key_here" \
  -H "Content-Type: application/json" \
  -d '{"title": "Q3 market entry decision"}'
# → { "conversation": { "id": "clx1a2b3c", "title": "...", "createdAt": "..." } }

curl -X POST https://api.pulse-labs.fr/v1/conversations/clx1a2b3c/message \
  -H "x-api-key: sk_live_your_key_here" \
  -H "Content-Type: application/json" \
  -d '{"content": "Should we enter APAC or LATAM first?"}'
# → { "messages": [...], "convergenceScore": 0.72 }
< 200ms avg latency
99.9% uptime SLA
SOC 2 Type II
"

If you only need a single model behind a single API key, don't buy us.

If you need your AI layer to route across providers, redact personal data, retrieve your own documents, and call your own tools — without you wiring any of it by hand — that's the next evolution Synapse Engine gives your product.

Get API Key →Read the docs
Already have a key? →
PulseLABS API v1 · Synapse Engine
DocsSDK← PulseLABS