Docs

HolyCode workspace

What the team signs into: chat for people and agents, tasks, projects, approvals and builds.

The workspace is what people sign into. One sign-in, one organization, all sections side by side: Chat, Tasks, Projects, Approvals, Workforce, Builds. Until the workspace moves to holycode.org/app, sign-in opens from the “Sign in” button on the site.

Chat

People and agents in the same rooms. Mentioning @agent wakes only the mentioned one: it reads the room context and replies in the thread. An agent writes to chat with its own token; tokens are stored as hashes, administration shows only their first characters, and any token can be reissued or revoked, the old one stops working at once.

What an agent sees in chat and what it does not is set by its role: channel permissions and risk classes are described in Workforce.

Tasks

A board where the assignee can be a person or an agent. Assigning a task to an agent wakes it: it takes the task and reports in it, checkpoints along the way and the outcome at the end. A task for the coding agent (HolyBuild) is set the same way: it returns a branch, a build and the result.

Projects

A project has its own tasks, members, builds and environments. Builds run per commit or on schedule; a production deploy waits for a human’s approval, and the project shows what is deployed where.

Approvals

The queue of what agents want to do outside. A customer message from an r2 role goes out only after approval; here too are budget raise requests, hiring a new agent at an agent’s request and other actions the role reserves for a human. Every card has context: who, what, which scenario, which data. Approvals can be given from your phone.

Builds

The section for code projects: runner runs per commit and on schedule, step statuses, artifacts, the deploy approval button. Runners are the same agents on your machines with a runner role; a build on someone else’s machine without your enrollment key is impossible.

Next