Rensei Platform Documentation

Rensei is the platform for running fleets of AI agents across your software development lifecycle, from interactive chat to scheduled, policy-governed workflows. This is the documentation for the paid platform. The open-source runtime (donmai) is documented at donmai.dev/docs.

Pick your path

Quickstart

Connect Linear and run your first SDLC agent in under 10 minutes.

Key Concepts

Projects, workflows, agents, sessions, and the layered model.

Install CLI

brew install rensei, authenticate, and set up compute.

How Rensei Works

The six-layer execution model - what the platform adds over the OSS runtime.

Workflow Editor

Design agent workflows on the visual React Flow canvas.

Default SDLC

The built-in five-phase workflow: research → development → QA → acceptance.

Sessions

Real-time activity streams, inspector, and Linear Agent integration.

API Reference

Worker protocol, public session API, webhooks, and auth.

Add a Sandbox Provider

Bring your own execution environment - Docker, K8s, E2B, and more.

How Rensei fits together

Rensei runs your agents on the execution substrate you choose - your own machines via the local daemon, or managed sandboxes (Docker, Kubernetes, E2B, Modal, Daytona, Vercel) - while the platform handles routing, governance, memory, and analytics.

By persona

Start with the Quickstart to see an agent run end-to-end. Then:

Worker protocol - register your own compute
Session activities API - stream agent output programmatically
Expression evaluator - {{ }} interpolation in workflow nodes
Agent cards - author and extend agents
CLI reference - full command catalog

Start with the Workflow editor canvas. Then:

Default SDLC template - the built-in workflow, ready to customise
Node palette - browse available nodes by integration
Gates - human approval, timer, and webhook gates
Workflow versioning - snapshots, diffs, and rollback
Workflow environments - staging vs production resource bindings

Start with the Org hierarchy. Then:

Model profiles and routing - scope cascade, work-type routing
Cedar policies - fine-grained access control
Audit trail - hash-chain, TSA anchoring, SIEM export
Capacity pools - substrate resolution and routing policy
Required workflow policies - seed-and-enforce for compliance

Start with Add a sandbox provider. Then:

Capacity pools - pool registration and substrate resolution
Worker protocol registration - the dial-in handshake
Auth modes - byok, metered, shared, host-session, local
Credential socket protocol - how credentials reach agents

Start with Interactive chat. Then:

Session list - find and filter running sessions
Session detail - the activity stream and health indicators
Linear Agent session - the Linear-native thread lifecycle

These docs are under active development. Pages marked beta or coming soon describe capabilities that exist in the product but are still being hardened for general availability.