glossary
The agent control plane, defined.
The vocabulary of running AI coding agents as a team: what the terms mean, why they matter, and how they fit together. Each definition stands on its own; each page goes deeper.
Agent control plane
An agent control plane is the management layer that lets a team run, coordinate, and supervise its AI coding agents from one place: shared context, scheduled runs, budgets, approvals, and an audit trail, across every tool and machine the team uses.
[read more →]Managed agent runs
A managed agent run is a coding-agent task executed under management: scheduled or triggered by the team, run on the team's own machine or CI, bounded by a hard token and time budget, and reported back with its status, cost, and a verifiable result.
[read more →]Guarded autopilot
Guarded autopilot is an execution policy for autonomous coding agents: tasks run without a human watching, but only inside hard guardrails: an allowlist of who may author them, an injection scan before any prompt is assembled, live budget enforcement, and a run prompt that forbids irreversible actions so the agent stops and describes them instead of executing.
[read more →]Interface contract (for agents)
An interface contract is a declared, versioned description of an interface one agent owns and other agents depend on, an endpoint shape, a shared type, a module boundary, published before the code is written so consumers build against the declaration instead of a guess.
[read more →]Contract negotiation
Contract negotiation is the process by which one agent proposes a change to an interface other agents depend on, the dependents respond with acceptance, requested changes, or a counter-shape, and the change applies only once reconciled, before any conflicting code is committed.
[read more →]Collision alert (pre-merge)
A collision alert is a warning delivered to an agent the moment an interface it depends on changes, while the agent is still writing code against the old shape, before either side has committed.
[read more →]Agent coordination layer
An agent coordination layer is shared infrastructure that lets multiple AI coding agents, across developers, machines, and tools, work on one codebase without duplicating work or breaking each other's interfaces, by sharing intent, contracts, claims, and decisions in real time.
[read more →]Approvals inbox
An approvals inbox is the human-in-the-loop surface of an agent control plane: the single place where decisions agents may not take alone, breaking interface changes and the questions agents raise, wait for a person to approve, reject, or answer.
[read more →]Agent run budget
An agent run budget is a hard limit, tokens, cost, iterations, or wall-clock time, attached to an autonomous agent task and enforced live by the runner, which kills the run when the limit is crossed rather than warning about it afterwards.
[read more →]Team memory (for agents)
Team memory is the durable, shared record of a team's decisions, conventions, briefs, and verifications that every agent, on any tool, machine, or later session, inherits automatically, so knowledge survives context windows and team turnover, human or agent.
[read more →]Prompt-injection scan
A prompt-injection scan is a check run over any text an agent is about to execute or ingest, a task description, shared content, a transcript, that flags instruction-override, destructive-command, secret-exfiltration, and jailbreak patterns before the text ever reaches the agent's prompt.
[read more →]