Skip to content

Two Modes of Operation

Sequant offers two ways to work: interactive in Claude Code chat, or autonomous via headless CLI.

AspectInteractiveAutonomous
InterfaceClaude Code chatTerminal CLI
ControlReview each phaseHands-off
Best forComplex issues, learningBatch processing
IterationManual /loopAutomatic quality loop

Use slash commands in Claude Code chat for step-by-step control.

Terminal window
/spec 123 # Plan implementation
/exec 123 # Build in worktree
/test 123 # UI verification
/qa 123 # Quality review
/loop 123 # Fix and retry
# Or all-in-one
/fullsolve 123
  • Learning Sequant — See each phase’s output
  • Complex issues — Review plans before implementation
  • Sensitive changes — Need human oversight at each step
  • Debugging — Investigate issues between phases
You: /spec 42
Claude: I've analyzed issue #42 and drafted a plan...
[Shows implementation plan]
You: Looks good, but let's use a different approach for the database query.
Claude: Updated plan. Ready for /exec?
You: /exec 42
Claude: Implemented in worktree feature/42-add-caching...
[Shows changes and test results]
You: /qa 42
Claude: QA complete. All AC items satisfied. Ready to merge.
  • Full visibility into each phase
  • Can intervene and redirect
  • Better for learning and understanding
  • Collaborative decision-making

Use the CLI for batch processing without interaction.

Terminal window
# Single issue
npx sequant run 123
# Multiple issues in parallel
npx sequant run 1 2 3
# Sequential processing
npx sequant run 1 2 3 --sequential
# With quality loop (auto-fix)
npx sequant run 123 --quality-loop
# Custom phases
npx sequant run 123 --phases exec,qa
# Preview without execution
npx sequant run 123 --dry-run
  • Batch processing — Multiple issues at once
  • CI/CD integration — Automated workflows
  • Simple issues — Bug fixes, docs updates
  • Overnight runs — Queue work and review later
Terminal window
# Process sprint backlog overnight
npx sequant run 45 46 47 48 --quality-loop
# In the morning, review PRs
gh pr list

Autonomous mode can automatically fix issues:

Terminal window
npx sequant run 123 --quality-loop --max-iterations 5

How it works:

  1. Runs phases (spec → exec → qa)
  2. If QA fails, runs /loop to fix
  3. Re-runs failed phases
  4. Repeats up to N iterations

Quality loop auto-enables for issues with these labels:

  • complex
  • refactor
  • breaking
  • major
  • Process multiple issues efficiently
  • No manual intervention needed
  • Consistent execution
  • Good for overnight batch work

Combine modes for the best of both:

Terminal window
# Start with interactive spec
/spec 123
# Review and approve plan
# Run implementation autonomously
npx sequant run 123 --phases exec,qa --quality-loop
# Review results in chat
/assess 123
ScenarioRecommended Mode
First time using SequantInteractive
Complex feature with unclear requirementsInteractive
Simple bug fixAutonomous
Sprint batch processingAutonomous
Security-sensitive changeInteractive
Documentation updatesAutonomous
Refactoring with breaking changesInteractive (spec), then autonomous

Configure in .sequant/settings.json:

{
"run": {
"qualityLoop": false,
"maxIterations": 3,
"sequential": false
}
}
Terminal window
# Override settings with flags
npx sequant run 123 --quality-loop --max-iterations 5 --sequential

CLI flags always override settings file values.