The CLI

Storybloq CLI.

Project memory and workflow in your terminal.

Give Claude Code and Codex a shared project record and a workflow: choose useful work, review the plan, preserve evidence, and continue across sessions.

Start a session

Load your project context.

Run /story in Claude Code chat, or $story in Codex chat, to load the project before you start working.

  1. i.Loads project status: stories, issues, phases, blockers
  2. ii.Shows what changed since last snapshot
  3. iii.Reads recent session handovers for continuity
  4. iv.Loads development rules and lessons learned
  5. v.Suggests the next story to work on

CLI + MCP

Project tools your agent can use.

Your coding client writes and runs the code. Storybloq gives it a project record and a workflow to follow.

Choose useful work

Recommendations explain what needs attention: blockers, unfinished stories, and work that would unblock others. Consult dependencies, recorded decisions, and lessons before choosing the next step.

Guide implementation

Take a scoped story through planning, independent review, tests, and completion. Your configured backends supply the reviews; the agent follows the workflow and reports progress.

Keep the evidence

Preserve approved plans and review findings. In gated workflows, acknowledgments can refer to a specific plan or staged code state, so you can inspect what was accepted.

Coordinate and continue

Share work across supported agent sessions, preserve assignments and handovers, and recover eligible interrupted sessions with ownership checks. Execution and recovery depend on your client.

See what was approved. Know what was checked. Evidence records review activity; it does not guarantee correct code. Keep tests, CI, and release checks in the workflow.

More workflow, when you need it.

Start with stories and handovers. Add the workflow that fits the work.

Autonomous mode: one story through completion

Use /story auto in Claude Code or $story auto in Codex for scoped execution through planning, review, implementation, and finalization. Your client needs the configured review tools and permission to run the project’s commands. FINALIZE can commit before COMPLETE.

Tiered orchestration: work through a backlog

The orchestrate skill organizes work into prepared specifications, implementation, verification, and review gates. Where the client supports model selection and subagents, stronger models handle judgment and cheaper tiers handle bounded implementation. Available dispatch and completion tools determine how the workflow runs; cost savings are workload-dependent.

Manager and worker: coordinate separate sessions

Duet mode records participants, scoped assignments, progress, and gate acknowledgments. These records survive the individual chats. Launching a worker, sending messages, and waiting for it require client capabilities or a configured transport; an arrangement record alone does not provide those operations.

Team mode: share the project record through Git

Enable team mode and install its merge driver in each checkout. Structured merges, work claims, and reconciliation help people and their agents share the ledger. Unresolved conflicts and duplicate display IDs still need attention; team mode does not eliminate every conflict.

Bus: exchange local peer messages

Connect two task-bound endpoints in one checkout with storybloq bus setup. Polling works explicitly; live delivery needs supported hooks or transport. Messages are peer advice, not owner approval. The optional idle-wake tier is for supported Codex CLI endpoints with a reachable compatible app-server and ownership checks. It cannot wake Codex Desktop; a wake request is not proof the peer read the message.

Recovery: continue from recorded progress

Handovers and saved session state support continuation in Claude Code and Codex. Claude usage-limit recovery is separate: ordinary sessions notify by default, while eligible autonomous sessions can resume after ownership, Git HEAD, and permission checks. Mid-FINALIZE stops require manual recovery. Codex does not have equivalent usage-limit auto-resume.

Run storybloq reference for command and tool arguments, or read the settings guide.

File convention

The .story/ directory.

Stories are saved as ticket records in tickets/. CLI commands keep the name ticket. Legacy files such as T-001.json coexist with canonical t-*.json files carrying a displayId. Stage the changed files Git reports; do not infer paths from display IDs.

config.jsonProject config (name, type, language)
roadmap.jsonPhase ordering and metadata
tickets/Your stories, stored as ticket records
issues/One JSON record per bug or gap
notes/Brainstorming and recorded decisions
handovers/Session continuity documents (markdown)
snapshots/State snapshots for session diffs (gitignored)

Pattern

Working across multiple repos.

The CLI walks up the directory tree to find .story/, so a single shared .story/ at a wrapper directory can serve many sub-repos. For projects that already have separate ledgers, federation connects their dependencies and recommendations while each repo keeps its history.

Read the multi-repo pattern guide