Documentation

Glossary

Every term the Coroid interface uses, defined once.

Terms are defined here once and linked from everywhere else.

Acceptance criteria

The checkable statements in a specification that decide whether work succeeded. Verification tests against them.

Advisory gate

A quality gate whose failure is recorded as a finding but does not stop the task. Contrast with a blocking gate.

Agent

A specialised worker that performs one stage of a task. Each agent has a fixed role and a restricted set of tools. See Agent roles.

Agent-hour

A unit of metered execution time. Counts while an agent is actively working a task, not while the task waits in the queue.

Agent slot

The unit of parallelism. One slot runs one task at a time. See Capacity and agent slots.

AI profile

A named model configuration assigned to an agent, trading cost, speed, capability and constraints such as data residency. See AI profiles.

Automation

Work that starts without anyone asking for it each time: a schedule, or a hook that runs when something happens. Automations are listed in the factory. See Automations.

Baseline

The base branch commit a plan started from. Before the plan's pull request merges, Coroid checks that the base branch has not changed since. If it has, a person accepts the new state or Coroid merges it in and re-verifies. See When the base branch moves.

Blocking gate

A quality gate whose failure stops the task until it passes or is overridden.

BYOK

"Bring your own key" — supplying your own model provider credentials so usage is billed by your provider at your rates rather than through Coroid.

Check

A single verification performed against a change. Checks are grouped into policies and run as gates.

Canary

A Quality Graph check run after a confirmed deployment to assess environment health. Its failure does not erase the successful deployment record. See Releases.

Coordinator

The agent that steers a task in Coordinator mode. It is the Architect, kept on the task after planning: it gives the developer, QA and reviewer their instructions one at a time, answers their questions and decides what happens after each report. It cannot overrule a verdict or finish a task that QA has not passed. See Agent roles.

Deployment attempt

One request or observed run for a release commit in one environment. Retries create another attempt; the release retains the earlier outcome. See Releases.

Discovery

Something Coroid found in your codebase while analysing it, which may become suggested work.

Finding

An issue recorded against a change — by a reviewer agent, a scanner, or an advisory gate. Findings are either fixed as rework or attached to the pull request.

Globalization

Making a product ready for every market: code that works in any language and culture (internationalization), then its translations (localization). Also the main-menu destination for both, and the Quality check that keeps every change world-ready. See Globalization.

Hook

An organization-level automation that reacts to a platform event. See Hooks.

Intake

The stage where a request becomes trackable work. A person files it through New Work, an issue import or a Prepare Work chat; another system files it through the work intake API or an event automation, such as one that watches Slack or Telegram. These requests pass through work qualification before any work is created. In Assistant mode, the chat assistant creates work directly, and so do the MCP server and Coroid's own automation. See Work qualification and Ways to start work.

Lane

The branch a plan's tasks integrate into. Tasks target the lane rather than your base branch, and the assembled result is promoted as a single pull request. See Tasks, plans, sprints and lanes.

Manual attestation

An Owner or Admin's record that an external pipeline deployed the pinned commit, with time and supporting reason or evidence. It records an observed deployment; it does not run CI or retroactively pass a gate.

MCP

Model Context Protocol. Coroid both connects to external MCP servers and can be used as one. See the Integrations overview.

Assistants such as Hermes, Codex and Claude Code use Coroid's MCP server to work with Coroid. See Coroid as an MCP server.

Plan

An ordered set of tasks grouped into phases, approved before code is written.

Policy

A named, versioned set of quality gates attached to projects.

Project

One repository plus its configuration, context and settings. The container work runs in.

Promotion

Opening the pull request that takes a lane's integrated work to your base branch, once it qualifies.

Provider key

A credential for a model provider that you supply. Encrypted at rest, never returned by the API, never included in settings exports.

Pseudo-localization

Rendering the product with a fake translation, accented, longer and reversed for right-to-left languages, to find text the catalogue missed, labels cut off and layout that does not mirror. See Pseudo-localization.

Quality gate

A check work must pass before progressing. See Quality gates.

Release

One pinned project commit followed through its environments, with checks, approval, deployment, and health recorded for each stop. A release may be versionless. See Releases.

Release gate snapshot

An immutable readiness result bound to a release commit, scope revision, environment, and policy version. Approvals refer to that result and become stale when the candidate or policy changes.

Release train

An ordered view of releases from several projects. It can show partial progress but does not promise one atomic deployment across repositories.

Review evidence

Artifacts attached to a pull request as proof of verification — preview deployments, scan results, coverage reports, test runs. See Review evidence.

Review rule

An organization-defined instruction that shapes what the reviewer agent looks for.

Rework

Sending a task back to the developer agent with findings, keeping its context, rather than starting over.

Sentinel

An earlier name for automations that watch for a condition. There is no separate Sentinel feature. See Automations.

Service account

A non-human identity for programmatic access, with its own scoped credentials.

Other systems use a service account to file requests through the work intake API. See From your own systems.

A qualified task from a service account waits for someone to start it, unless an owner or admin set the account to start qualified work automatically.

Specification

The statement of scope, exclusions and acceptance criteria produced before work is built.

Sprint

A time-boxed container for work, with its own metrics and timeline.

Task

One unit of work that ends in one pull request. The unit that actually executes.

Work qualification

The checks a request passes before any work is created. Qualification screens the request for sensitive values, checks it against the current release, looks for duplicates, verifies or reproduces what it can, and settles whether it becomes a task, a plan, a report or a schedule. Requests that wait on a decision are listed under Factory → Review.

Workspace

The isolated, disposable environment where an agent clones your repository and works. One per task.

World readiness

What the code on a project's default branch still assumes about one language or one culture, with fix tasks for it. See World readiness and Culture.