Docs
Glossary
The documentation terms in one line each, from organization and region to the company package.
- HolyCode: the platform and the workspace in the cloud that people sign into.
- HolyAgent: the agent on your machines: models on your keys, channels, memory, skills.
- HolyBuild: the coding agent: takes tasks from HolyCode, returns a branch and a build.
- Organization: a team of people and agents with one sign-in and one set of data.
- Region: where organization data lives: US (worldwide) or RU (Russia and Asia); chosen at creation, never changes.
- Workspace: the sections for people: Chat, Tasks, Projects, Approvals, Workforce, Builds.
- Workforce: the section with roles, teams, scenarios, budgets and nodes.
- Role: a document on what the agent does, what is forbidden and which channels it may use; versioned.
- Risk class: r0 reads only, r1 works with data without writing outside, r2 writes to people after approval.
- Team: a set of roles and scenarios for one business job; seven ready-made in the catalog.
- Scenario: a graph of steps in the
hpl/v1format that the agent follows for a task. - Task: one run of a scenario; it has checkpoints and a run transcript.
- Checkpoint: the agent’s note along the way: what is done, what is next.
- Run transcript: what the task was started with, what the agent asked the model, what it answered.
- Approval: a card in the queue where a human decides: an outbound message, a budget, a hire, a deploy.
- Budget: a per-task limit (enforced by the server) and a monthly per-agent limit (80 % warning, 100 % pause).
- Pause: the agent state in which no new tasks are handed out; set by an operator or by the budget.
- Node: one HolyAgent on one machine, with a version, a status and a manager.
- Enrollment key: a one-time key to connect a node; after the first run the agent keeps a permanent key.
- Access key: a key for scripts and external systems with
readorread, writerights. - Connector: a channel or integration a scenario needs; a task goes only to a node that has it.
- Channel: WhatsApp, Telegram, Instagram or voice connected on the agent’s machine.
- Skill: a folder with
SKILL.md: what the agent can do and how. - Company package: roles, scenarios and skills in one file for transfer and for keeping in git.
- Escalation: a stuck task or an uncertain answer moving up the chain of managers or to a human.
- Runner: an agent with a build role for code projects.