From idea to shipped code,
eight stages, zero babysitting.
Discovery, interview, planning, execution, audit, gap-fill, final review, commit. Tendril runs the full pipeline — you review and approve.
Discover & Plan
Discovery → Interview → Planning — your project gets mapped and broken into tasks
Point Tendril at your project folder. It reads your codebase structure, asks clarifying questions about your goals, then generates a prioritized plan of subtasks on your Kanban board.
- 1. Discovery — Tendril scans your project files and builds a structural map in seconds
- 2. Interview — your main agent asks targeted questions about what you want built
- 3. Planning — work is broken into prioritized subtasks (P1 runs first, then P2, then P3)
- Pro: Knowledge Graph primes agents with patterns and memory from past sessions for higher-quality plans
- You review and approve the plan before any agent starts executing
- All data stays local — nothing is sent to Tendril servers
Execute & Audit
Agents execute in parallel → auto-scored → gap-fill retries if needed
Sub-agents execute subtasks in parallel, each working in an isolated git worktree. Pro users get automatic quality scoring after each subtask — below-threshold work gets retried with targeted gap-fill instructions.
- 3. Execution — P1 subtasks run in parallel, then P2, then P3. Each agent works in isolation.
- 4. Audit (Pro) — every subtask is auto-scored. Work that passes ships to review.
- 5. Gap-fill (Pro) — below-threshold subtasks get retried with specific fix instructions (up to 3 retries)
- Free: 2 parallel sub-agents. Pro: unlimited (hardware-limited).
- Live status dots show which tasks are actively being worked on
- Pro: execution levers (error memory, pattern extraction, scope pre-read) cut costs 8-42%
Review & Commit
Final review → your approval → merge — and the Knowledge Graph learns
A senior-level final review checks the whole project. You review the full diff and approve before anything merges. Every decision and pattern gets saved to project memory for next time.
- 6. Final Review (Pro) — a senior-level check evaluates the complete output holistically
- 7. Your Approval — you see the full diff with a clear summary. Nothing merges without sign-off.
- 8. Commit — approved changes merge cleanly into your branch
- Pro: Knowledge Graph saves every decision, pattern, and architectural choice
- Each session gets cheaper — costs drop from $1.05 to $0.50/subtask over 50 sessions
- Cross-project learning: patterns from one project help all your others
Discover & Plan
Discovery → Interview → Planning — your project gets mapped and broken into tasks
Point Tendril at your project folder. It reads your codebase structure, asks clarifying questions about your goals, then generates a prioritized plan of subtasks on your Kanban board.
- 1. Discovery — Tendril scans your project files and builds a structural map in seconds
- 2. Interview — your main agent asks targeted questions about what you want built
- 3. Planning — work is broken into prioritized subtasks (P1 runs first, then P2, then P3)
- Pro: Knowledge Graph primes agents with patterns and memory from past sessions for higher-quality plans
- You review and approve the plan before any agent starts executing
- All data stays local — nothing is sent to Tendril servers
Execute & Audit
Agents execute in parallel → auto-scored → gap-fill retries if needed
Sub-agents execute subtasks in parallel, each working in an isolated git worktree. Pro users get automatic quality scoring after each subtask — below-threshold work gets retried with targeted gap-fill instructions.
- 3. Execution — P1 subtasks run in parallel, then P2, then P3. Each agent works in isolation.
- 4. Audit (Pro) — every subtask is auto-scored. Work that passes ships to review.
- 5. Gap-fill (Pro) — below-threshold subtasks get retried with specific fix instructions (up to 3 retries)
- Free: 2 parallel sub-agents. Pro: unlimited (hardware-limited).
- Live status dots show which tasks are actively being worked on
- Pro: execution levers (error memory, pattern extraction, scope pre-read) cut costs 8-42%
Review & Commit
Final review → your approval → merge — and the Knowledge Graph learns
A senior-level final review checks the whole project. You review the full diff and approve before anything merges. Every decision and pattern gets saved to project memory for next time.
- 6. Final Review (Pro) — a senior-level check evaluates the complete output holistically
- 7. Your Approval — you see the full diff with a clear summary. Nothing merges without sign-off.
- 8. Commit — approved changes merge cleanly into your branch
- Pro: Knowledge Graph saves every decision, pattern, and architectural choice
- Each session gets cheaper — costs drop from $1.05 to $0.50/subtask over 50 sessions
- Cross-project learning: patterns from one project help all your others
See your agents working in parallel
Tendril surfaces every project, agent, and cost as a live graph — so you always know what's running, what's waiting, and what it's costing you.
Ready to let agents do the heavy lifting?
Start free. No credit card. Bring your own API key.
Frequently Asked Questions
Everything you need to know about how Tendril works under the hood.