Skip to content

Glossary

Alphabetical reference of Colony terminology. Each definition links back to the Core Concepts page where that concept is introduced in context.


A short disclosure footer Colony appends to most of the issue comments, PR descriptions, review bodies, and created issue bodies it writes. Enabled by default to support EU AI Act Article 50(1) transparency. Some pull request comments, such as Mayor replies, escalations and close notices, carry Colony’s signature instead of the footer, and an approval on a recorded verdict states that Colony made it. The footer text and opt-out behavior are configured under attribution in colony.config.yaml. Have counsel review this wording before the page goes public.


One of the four human roles in a Colony deployment. The Author files issues, writes acceptance criteria, and validates the result. Typically product managers, product owners, or tech leads briefing work for their teams.


A function that implements one phase of the pipeline for a custom workflow. Executors receive a task from the Worker, perform their work (typically via Claude Code), and return an ExecutorOutcome whose outcome is one of the outcome slugs the executor declared with defineExecutor. The workflow definition’s per-state on: map, not the executor, decides the next state. See the Pipeline States reference for the built-in states.


A discrete piece of work — a GitHub issue, an Azure DevOps work item on self-hosted Colony, or a similar ticket. Issues are Colony’s primary input: they describe what needs to be done, and the pipeline drives them to completion autonomously. See Core Concepts — Issues for what makes a well-formed issue.


The projection of Colony’s pipeline state onto your version control system. Colony writes a label (e.g. colony:ready-for-dev) to every GitHub issue to reflect its current pipeline state. State labels are a projection: removing or changing one in GitHub does not move the pipeline. The exceptions are the control labels Colony reads as incoming commands: colony:enqueue adds an untracked issue to the pipeline, colony:paused pauses or resumes an issue, and colony:ignore keeps an untracked issue out of automatic intake. See Core Concepts — Labels and Tags.


The central conductor of the pipeline — Colony’s always-on pipeline orchestrator. In Colony Cloud there is one Mayor per tenant. A self-hosted deployment runs one Mayor, which serves every tenant in its configuration. It ingests newly created issues, delegates work to available Workers, manages cross-issue dependencies, and runs recovery protocols when work stalls. See Agents — Mayor.


One of the four human roles in a Colony deployment. The Operator configures the colony, allocates Worker capacity, and intervenes when work stalls — typically by posting slash commands on blocked issues. Typically platform engineers, tech leads, or whoever owns the deployment.


Colony’s model of your development process. A deterministic state machine that drives issues from New (first ingested) to Done (PR merged). Each tenant has exactly one pipeline. The pipeline is opinionated at the state-machine level but configurable at the behavior level. See Core Concepts — Pipeline.


The data layer that backs every pipeline. The authoritative record of all issues, their current states, work history, agent transcripts, cost events, and audit logs. Postgres-backed. GitHub state labels are a projection of Pipeline Store state; editing them does not drive it, apart from the colony:enqueue, colony:paused and colony:ignore controls. See Core Concepts — Pipeline Store.


A codebase hosted in a version control system: GitHub, or Azure DevOps on self-hosted Colony. Colony Cloud connects to GitHub only. Each repository belongs to exactly one tenant. Colony reads configuration, file structure, and history from the repository and writes code, opens pull requests, and merges changes back into it. See Core Concepts — Repository.


One of the four human roles in a Colony deployment. The Reviewer judges Colony’s pull requests for architectural fit and intent — typically senior ICs, tech leads, or anyone who would review a colleague’s PR today.


A Colony feature that allows Colony to file issues against its own configured repositories based on observed patterns, test failures, or configured tracks. Self-improvement issues flow through the pipeline like any other issue. Tracks are managed in the Colony Cloud dashboard or through the API; self_improvement.tracks entries in colony.config.yaml are deprecated and ignored at runtime. See Self-Improvement Tracks.


Commands posted as comments on GitHub issues to control pipeline state. Operators post them to intervene when work stalls or to trigger specific behaviors. Common examples:

CommandEffect
/colony:retryUnblock a failure-blocked issue and re-queue it
/colony:reimplementClose the current PR and re-queue the issue for a fresh development attempt
/colony:state <state>Move the issue to a specific pipeline state
/colony:reviewRe-trigger the deterministic checks and the LLM review on a pipeline PR; also works on external (human-authored) PRs
/colony:pausePause a running issue
/colony:resumeResume a paused issue

See Reference: Slash Commands for the complete catalog, exact syntax, and per-command state restrictions.


One of the four human roles in a Colony deployment. The Sponsor charters the rollout, sets autonomy grants, and tracks throughput and cost at an organizational level — typically engineering managers, directors, or whoever owns the budget.


The current position of an issue in the pipeline. Every issue is in exactly one state at any time. The state determines what work happens next and which executor will claim it. The built-in colony-default workflow defines 16 states, and custom workflows may define others; see the full Pipeline States reference for the complete list, valid transitions, and the state machine diagram.


Colony’s top-level organizational unit and a configuration boundary. Every resource inside a tenant (repositories, Workers, API keys, GitHub access) is shared across everything within that tenant. Colony scopes pipeline data, Workers and credentials by tenant. Use a single tenant for multi-repo projects that share resources, and separate tenants to keep independent groups’ work, credentials and budgets apart. Each customer organization runs on its own Pipeline Store, and that store is the isolation boundary between customers. See Core Concepts — Tenant.


Self-improvement tracks are configured prompts that tell Colony what kinds of improvements to look for and file as issues on a regular cadence. Each track defines a prompt, a cooldown period, and the target repository. Tracks are managed in the Colony Cloud dashboard or through /api/repos/{owner}/{repo}/si-tracks. self_improvement.tracks entries in colony.config.yaml are deprecated and ignored at runtime, and colony config migrate removes stale entries. See Self-Improvement Tracks.


An individual unit of work capacity — a container running a Colony agent process. Workers listen for tasks enqueued by the Mayor, claim one task at a time, execute the work autonomously, and report the result back to the Pipeline Store. Any Worker can handle any task type (analyze, plan, develop, review, merge). Pool size controls how many issues are actively worked in parallel. See Agents — Workers.


The five main task types that Workers execute, one for each phase of the pipeline:

Work typeWhen it runs
AnalyzeTriages the issue — estimates complexity, produces an implementation plan, transitions to ready-for-dev or flags for clarification.
PlanDecomposes an epic-scale issue into smaller, actionable subtasks.
DevelopImplements the issue — writes code, runs review checks, and opens a pull request.
ReviewInspects the pull request — runs deterministic checks, performs an LLM review, and either approves or requests changes.
MergeRebases and merges an approved pull request into the target branch.

See Agents — Work Types.