Temper documentation

Temper is an event-sourced coordination substrate for agent-assisted development. It keeps
goals, tasks, sessions, research and decisions durable across conversations, so understanding
compounds instead of being rebuilt every time.

These pages are the operational half: how to install it, run it, and integrate with it. The
why — attention as the organizing constraint, what a cognitive map is, what the substrate
guarantees — lives at temperkb.io.

Three ways in

Pick the one that matches what you're doing. Each is a route, not a wall — they cross-link
freely.

Using Temper — you want Temper working for you and your agents:
install the CLI, wire up Claude, keep a session's context alive, grow a cognitive map.

Running Temper — you're standing up a deployment: self-host
it, wire an identity provider, bootstrap an org, connect Slack and GitHub, get telemetry out.

Building against Temper — you're writing code that talks to
Temper: the HTTP API, machine credentials, and the language SDKs.

Reference

Generated, not written. Each of these is emitted from the thing it describes and gated in CI,
so a page that disagrees with the software is a defect rather than a stale document.

CLI reference — every temper command, emitted by walking
the built binary's own --help.

Configuration reference — every field of TemperConfig,
with its type, its real default, and the doc comment from the struct.

Every API endpoint and schema is generated from the router itself for the same reason, and is
published alongside these pages.

Temper documentation

Three ways in#

Reference#

Three ways in

Reference