Current workspace

Start with a clear picture of the work.

Perseus Context Engine reads the sources you choose and writes one briefing for the agent and the people reviewing it.

Without a briefing

The agent has to find the port, test state, and last decision in separate sources.

One briefing

API_PORT: 3001
Tests: inspect the current CI record
Checkpoint: review pending
Sources: 4 verified

How it works

Turn workspace details into one useful file.

1Choose sources

Name the files and services the workspace trusts.

2Build locally

Perseus reads them and applies the limits you set.

3Review the file

The agent starts from a briefing a person can inspect.

First briefing

Get started in three commands.

PyPI currently publishes `perseus-ctx` 1.0.26. The package root is the canonical current-version link.

pip install perseus-ctx==1.0.26
cd your-project && perseus quickstart
perseus render .perseus/context.md -o AGENTS.md

Permissions

Read-only is a starting point, not a security boundary.

The standard renderer reads local sources and writes only the output path you name. HTTP directives, network service modes, and shell commands are separate opt-ins and use the permissions of the account that runs them.

See the Context Engine security details →

Stores: No independent long-term memory. The briefing is a normal file.
Sends: No conversation or model prompt on the local render path.
You control: The sources, output path, and every optional network or execution surface.