The whole loop — in one window.
Every capability Tendril ships with, what it does, and why it matters. For the buying decision, see Pricing. For the data backing the claims, see the 120-trial A/B study.
Built Different. By Design.
Tendril isn't another chat window — it runs your whole project. AI agents that remember your codebase, a task board that updates itself, and your code that stays local by default.
Knowledge Graph that compounds
Tendril indexes your codebase and remembers decisions, patterns, and mistakes across every project you build. Lessons learned in one project prime planning in the next. Build five projects, and the sixth starts with the lessons of all five baked in — without writing a single prompt about them.
Multi-agent execution, in parallel
A master agent proposes a plan you review and edit. After approval, parallel sub-agents work different files simultaneously in isolated git worktrees. Capped only by your machine and your provider rate limits — no artificial throttle.
Kanban that fills itself
Backlog → In Progress → Done as agents work. Priority-based scheduling runs P1 first, then P2 in parallel, then P3. You watch what's happening; you're not constantly approving sub-decisions.
Live preview + annotation fix-flow
See your app render in a live iframe inside Tendril. Draw circles, arrows, and notes on the preview, type a comment, send it. Agents read your annotations directly and the fix lands as a card on the same kanban. The fastest "this looks wrong" → "this is fixed" loop in any AI dev tool.
Bring Your Own API Key
Use Anthropic, OpenAI, or Google. Or pass through your existing Claude Code subscription for one flat fee instead of per-token charges. AI costs go directly from your machine to the provider — Tendril never marks up compute.
Audit that catches what you would
Every finished project ends with an auto-scored audit and gap-fill retries on below-threshold work. A senior-level final review produces the project score, summary, and clickable improvement suggestions — fix cards that spawn agents to do the work with one click. Plus injection detection, scope enforcement, and command blocking under the hood.
Knowledge Graph that compounds
Tendril indexes your codebase and remembers decisions, patterns, and mistakes across every project you build. Lessons learned in one project prime planning in the next. Build five projects, and the sixth starts with the lessons of all five baked in — without writing a single prompt about them.
Multi-agent execution, in parallel
A master agent proposes a plan you review and edit. After approval, parallel sub-agents work different files simultaneously in isolated git worktrees. Capped only by your machine and your provider rate limits — no artificial throttle.
Kanban that fills itself
Backlog → In Progress → Done as agents work. Priority-based scheduling runs P1 first, then P2 in parallel, then P3. You watch what's happening; you're not constantly approving sub-decisions.
Live preview + annotation fix-flow
See your app render in a live iframe inside Tendril. Draw circles, arrows, and notes on the preview, type a comment, send it. Agents read your annotations directly and the fix lands as a card on the same kanban. The fastest "this looks wrong" → "this is fixed" loop in any AI dev tool.
Bring Your Own API Key
Use Anthropic, OpenAI, or Google. Or pass through your existing Claude Code subscription for one flat fee instead of per-token charges. AI costs go directly from your machine to the provider — Tendril never marks up compute.
Audit that catches what you would
Every finished project ends with an auto-scored audit and gap-fill retries on below-threshold work. A senior-level final review produces the project score, summary, and clickable improvement suggestions — fix cards that spawn agents to do the work with one click. Plus injection detection, scope enforcement, and command blocking under the hood.
AI Task Board
Your Kanban board fills itself. Tasks flow from Backlog → In Progress → Done as AI agents work. Priority scheduling runs P1 first, then P2/P3 in parallel.
Security Monitoring
35 injection patterns detected in real-time. Scope enforcement, command blocking, and budget caps keep agents safe. Every action is logged and auditable.
Quality Audits
ProEvery subtask is quality-scored by a senior-level AI reviewer. Critical issues are auto-fixed before code ships. Nothing ships without passing audit.
Plan Review
Review AI-generated plans before a single line of code is written. Approve, reject, or edit any subtask. You stay in control of the architecture.
Parallel Agents
Multiple AI agents work simultaneously on different subtasks. P1 tasks run first, then P2 and P3 in parallel. Hardware and API keys are the only limits.
Pro features marked above require a Pro subscription.
Ready to see the tiers?
Free gets the Medium and Complex pipelines, unlimited projects, and quick edits through the project chat. Starter adds Tendy chat (conversational multi-project orchestrator), live preview, Knowledge Graph, annotation, and 3 parallel agents. Pro adds clickable improvement suggestions on finished projects, cross-project memory, GitHub shipping, and unlimited parallelism. Founding pricing is locked for life.