Everything Cognipeer, in one place
Product guides for the teams who use Studio and Pulse, platform documentation for the people who run Console, and API references for everyone building on top of it.
- Self-host or hosted
- Console is free and open source
- TypeScript SDKs
- No vendor lock-in
Two products your team works in
Studio is where teams design, test, and ship AI agents. Pulse is the workspace assistant that runs the work and keeps people in control with tasks, approvals, memory, and files.
Cognipeer Studio
Build Peers, design Flows on a no-code canvas, connect data sources, add tools, publish to channels, and evaluate each version before it reaches users.
Cognipeer Pulse
A workspace assistant built around a persistent timeline. It runs tasks, asks for approvals, connects your tools, manages files and memory, and works by chat, voice, or webhook.
The control plane underneath it all
Console is the open-source AI control plane the platform runs on. It is documented for two readers at once — the person operating it, and the person building against its API.
Cognipeer Console
OpenAI-compatible LLM routing, a knowledge engine, an MCP hub, guardrails and PII policies, agent tracing, GPU fleet management, and cost optimization — running on your own infrastructure.
Build on Cognipeer
The Agent SDK is the runtime that powers Studio and Pulse, and it is open source. Around it sit SDKs for Console, the hosted API, the chat interface, and knowledge graphs.
@cognipeer/agent-server
Run agents as a service: REST endpoints, storage providers, auth, streaming, files, and background tasks.
@cognipeer/console-sdk
Typed client for every Console endpoint — chat, vectors, agents, tracing, guardrails, memory, and more.
@cognipeer/sdk
Talk to hosted Cognipeer from your app: conversations, flows, client tools, and structured output.
@cognipeer/chat-ui
React components for a production chat surface — streaming, tool calls, history, theming, uploads.
@cognipeer/observability
Ship agent traces from OpenAI Agents, LangChain, LangGraph, n8n, and OpenTelemetry into Console.
@cognipeer/graph-builder
Turn files and text into a queryable knowledge graph with semantic enrichment and incremental updates.
Pick the path that matches your work
Three ways into the platform, depending on whether you are evaluating it, running it, or building on it.
Create your first Peer
Configure an AI agent in Studio, connect a data source, add a tool, and chat with it. About five minutes.
Start in Studio →Stand up Console
Deploy the control plane, connect a provider, create your first model, and route an OpenAI-compatible client to it.
Read the Console guide →Write your first agent
Install the Agent SDK, give an agent tools, and run a loop with planning and approvals in your own code.
Open the Agent SDK →Latest releases
Recent changes across the platform. Each product keeps its own full changelog.
Cost & Optimization, Prompt Optimizer, and the model matrix
Spend attribution, the Analysis workbench, and automated Prescriptions moved to the enterprise edition, alongside model-switch recommendations backed by parity tests on real traffic.
Pulse becomes an app platform
Apps with their own surfaces, a manifest format, organization app management, and REST token scopes — the model for letting third parties into the timeline.
Client Tools, Webchat, and on-premise subscriptions
Client-side tool execution, Webchat improvements, on-premise subscription management, and an updated model lineup.