Use Cases
This page focuses on high-ROI, real-world CCCC workflows.
How to Read This Page
Each scenario includes:
- Goal
- Minimal setup
- Execution flow
- Success criteria
- Common failure points
Use Case 1: Builder + Reviewer Pair
Goal
Deliver one bounded change with an independent review and a clear owner for integration. Start with two agents before adding more roles.
Minimal Setup
This example uses members that CCCC starts. Install and sign in to the runtimes you intend to use first; follow the CLI Quick Start or Web UI Quick Start for the host setup. To connect sessions you already have open instead, use Native sessions.
For a project with no existing CCCC Group, register a pair:
cd /path/to/repo
cccc attach .
cccc actor add builder --runtime codex
cccc actor add reviewer --runtime claude
cccc actor listThese runtime choices are an example, not a required pairing. The first agent in a new Group is its foreman; here that is builder. If the project already has a Group, inspect its members and choose unused names instead of replacing existing agents.
Merge the short AGENTS.md template into the project's instructions, including its own commands and constraints. CCCC does not make every runtime load AGENTS.md; check the runtime's instruction loading before the first task.
Start the new Group's two members:
cccc group startWhen adding members to an existing Group, start only those members with their Web controls instead of starting every member in the Group. Before assigning work, complete any pending runtime onboarding or trust prompt in the Web confirmation terminal. See runtime interaction.
Agree where each member will edit. A shared Group or scope does not guarantee the same checkout: a managed Claude background session may enter a worktree. See where managed Claude edits files for configuration and cleanup. Do not assume a clean main checkout means another member has made no changes.
Execution Flow
- Give the builder one concrete behavior to change, acceptance cases and boundaries. Name the reviewer and integration owner; here the builder integrates after review. For example: "Remove duplicate IDs from the CSV export. Add coverage for repeated IDs and keep the public API unchanged."
- Ask the reviewer to inspect the requirements and current implementation independently. Agree file ownership before either edits; review can start with a read of the existing behavior while implementation proceeds.
- Keep work that needs a lasting owner or completion record in a CCCC task. Keep the goal and shared constraints in the brief, rather than copying every local todo or chat message onto the board.
- The builder hands off the actual checkout/branch, changed files, checks and remaining uncertainty. The reviewer examines that change and reports actionable findings with evidence, or says what was checked and found clear.
- The integration owner resolves confirmed findings and reruns affected checks. Report the result and remaining limits to the user; commits, pushes and releases follow the authority the user granted.
For a concrete reply obligation, the human can select Send + Reply in the Web UI, or use the CLI's --mode request-reply. Ordinary agent discussion follows CCCC's built-in Send/Mail guidance.
Use the next similar task to check whether an agreed workflow adjustment helped.
Success Criteria
- The agreed behavior and acceptance cases are satisfied.
- Review findings refer to the actual change and can be acted on.
- The result is integrated where intended; the handoff links acceptance checks and their results, with remaining limits stated.
Common Failure Points
- Task scope too broad.
- Reviewer lacks explicit acceptance criteria.
- Both agents edit the same files without agreeing ownership.
- A worktree contains the result but the integration owner checks only the main checkout.
- Runtime confirmation is still pending, or changed instruction files have not been read by the running sessions.
- A concrete review was requested without a reply obligation or clear handoff.
Use Case 2: Foreman-Led Multi-Agent Delivery
Goal
Split one medium project into parallel tracks while keeping alignment.
Minimal Setup
cccc actor add lead --runtime claude
cccc actor add frontend --runtime codex
cccc actor add backend --runtime grok
cccc actor add qa --runtime kimi
cccc group startExecution Flow
- Foreman records the current objective, focus and constraints in the shared brief, and uses tasks for work with durable owners and completion criteria.
- Assign focused tasks via direct recipients.
- Enforce checkpoint reminders through Automation rules.
- Foreman integrates and resolves conflicts.
- QA agent validates key acceptance criteria before handoff.
Success Criteria
- Parallel execution without major rework churn.
- Clear ownership per track.
- Traceable decision history in ledger.
Common Failure Points
- Missing shared architecture baseline.
- Agents editing same surfaces without ownership rules.
- No explicit integration checkpoints.
Use Case 3: Mobile Ops with IM Bridge
Goal
Operate long-running groups from phone while keeping a reliable audit trail.
Minimal Setup
cccc im set telegram --token-env TELEGRAM_BOT_TOKEN
cccc im startThen /subscribe in your IM chat.
Execution Flow
- Receive progress/error notifications in IM.
- Send escalation commands to
@foremanfrom mobile. - Switch to Web UI for deep debugging only when needed.
- Keep all critical decisions in CCCC messages (not only in IM thread).
Success Criteria
- You can intervene without laptop access.
- Critical context remains in ledger.
- Downtime is reduced for overnight or offsite ops.
Common Failure Points
- Exposing Web UI without proper token/gateway.
- Using IM as the only source of truth.
- No restart/recovery playbook.
Use Case 4: Repeatable Agent Benchmark Harness
Goal
Run comparable multi-agent sessions with stable logging and replayability.
Minimal Setup
- Define fixed task prompts and evaluation criteria.
- Use the same Copy Groups package and runtime setup per run.
- Keep automation policies deterministic.
Execution Flow
- Create a baseline group and export a Copy Groups package.
- Run multiple sessions with different runtime combinations.
- Collect ledger and terminal evidence.
- Evaluate outcome quality and operational stability.
Success Criteria
- Comparable runs with low setup variance.
- Reproducible evidence set (
ledger, state artifacts, logs). - Clear model/runtime tradeoff signals.
Common Failure Points
- Hidden prompt drift between runs.
- Uncontrolled environment differences.
- Missing run metadata in messages.
Recommended Next Reads
docs/guide/operations.mddocs/reference/positioning.mddocs/reference/features.md