Every PairCoder project ships with seven named agents. Here is who does what, and which plan turns on their automated dispatch. The agent files themselves are in every project on every plan; what Team adds is the automation around Vaivora and Divona.
Design and planning
Turns a goal into a plan. It clarifies requirements, weighs approaches and tradeoffs, and breaks the work into tasks with acceptance criteria. It reads your code but does not change it.
Dispatched by: Plan workflow: /pc-plan, or invoke it by name in Claude Code.
Implementation
Does the hands-on work for a task: writes tests first, makes them pass, runs the checks and records what changed. It works inside the limits the task and your enforcement gates set.
Dispatched by: Task workflow: /start-task; bpsai-pair engage runs it per task (Team).
Code review: quality and correctness
Reads a change for bugs, unclear logic, missing tests and departures from your project's conventions, then reports findings ranked by severity. It reviews and does not edit.
Dispatched by: bpsai-pair review on a PR or branch.
Security and compliance review
Looks for vulnerabilities, leaked secrets, risky dependencies and gaps against your compliance checklist. Like Nayru, it reports what it finds and leaves the fix to the driver.
Dispatched by: bpsai-pair review on a PR or branch.
Pre-execution security gate
Reviews a risky action before it happens. When Claude Code is about to run a command, commit code or open a PR, Bellona reads what is about to run and reports whether it is safe to proceed. It reviews; the mechanical blocking is done by the shipped hooks (secret scan, branch guard, filing gate), which run regardless.
Dispatched by: Claude Code consults it on its own before commands, commits and PRs that match its description; you can also ask for it by name.
Cross-module review: contract breaks and dependency ripple
Follows a change outward. It looks for broken contracts between modules, dependency conflicts and infrastructure effects that a file-by-file review would miss, and reports them at plan time and on review.
Dispatched by: bpsai-pair vaivora; engage's cross-module seam audit at plan time; added to bpsai-pair review only when the diff exceeds 500 changed lines or 10 files.
Browser QC
Runs your QC suites in a real browser, scenario by scenario, and hands back a structured pass or fail report with what it saw at each step.
Dispatched by: /run-qc in Claude Code, or the QC enforcement gate.
The review command always runs Nayru and Laverna on a PR or branch, and adds Vaivora only when the diff exceeds 500 changed lines or 10 files. Engage runs the driver for each task and Vaivora at plan time (Team). Divona runs through the /run-qc slash command in Claude Code or the QC enforcement gate (Team). The agent files live under .claude/agents/ in your project, so on any plan you can also invoke an agent by name inside Claude Code.
Fleet operations is the Enterprise feature for running PairCoder across a workspace of repositories. fleet check, fleet audit and fleet verify check, audit and verify every repository in the workspace, fleet check --upgrade upgrades stale ones, and fleet audit --bypasses reports where gates were bypassed. Team can buy it as an add-on.
The Computer package, bpsai-computer, is a computer that runs a fleet of these agents across an organization's repositories. Dispatch, review landing and fleet operations sit under one audited ledger. It is sold by BPS Star Fleet membership, not as a PairCoder tier, and PairCoder is included.