Skip to main content
Bridge
Core Concepts › Build Plan

Bridge v0.12.0

Build Plan

A Build Plan is what Bridge produces when you tap "Start a venture" on a constellation: a portable, markdown-only spec set that any coding agent can read and implement against.

What triggers a Build Plan

The trigger is the Start a venture button — on a note's detail screen, on a constellation popup in the graph, or in the Plan tab. It opens the venture picker, where the source notes come pre-checked and you can add or remove notes before generating. Tap Build Plan in the picker to run.

Bridge gathers context (the notes you selected, their expansions, the constellation theme, your user-behavior memory) and runs the generator in the background. It usually takes two or three minutes, and you don't need to keep the app open: Bridge sends a notification when the plan is ready, or if it fails. The plan shows in the Plan tab as generating the moment you tap, and one tap makes one plan, even if you tap twice or the network is slow.

The output

Bridge generates a spec set — markdown files designed to be agent-readable and human-readable. Each file has a strict purpose and a tight size budget so the whole set fits inside any modern coding agent's context window.

See Build Plan output files for the file-by-file breakdown.

Plus one CLI overlay file at the project root — AGENTS.md — that points your coding agent at the spec set in the right reading order.

The Plan page

Every Build Plan run lands in the Plan tab. Each artifact shows:

  • Which note triggered it.
  • When it was generated.
  • The files, viewable in-app.
  • Download options: zip and share sheet.

If a plan fails (for example, your AI provider's key ran out of credit), open it and tap Try again. That reruns the same plan rather than starting a new one. Its Inbox item follows it from started to ready or failed, and tapping it opens the plan.

You can regenerate a plan at any time. Start a venture from the same note again — the new plan uses your current context (latest notes, latest expansions). The old version stays in the Plan tab archive so you can choose which to use.

When a plan fails

A plan that fails says why on the plan, for example that your AI provider's key is out of credit, or "This Build Plan couldn't be generated. Try again." when there's no specific reason. When the cause is your key or model, it offers AI & Models settings. If your plan doesn't allow the Build Plan, Start a venture and Try again open the upgrade notice, which says why and offers the plan that includes it.

How the agent reads it

When you drop the plan files into your project directory, the AGENTS.md overlay at the project root tells your coding agent the reading order:

  1. 00_index.md orients the agent
  2. 07_scope.md captures the constraints early
  3. 01_brief.md provides what and why
  4. 03_architecture.md sets stack and structure
  5. 04_data_model.md defines pre-named structures
  6. 05_tasks.md is the work checklist

06_acceptance.md is consulted for "done" criteria. 02_user_flow.md is consulted for the demonstrable end-to-end flow.

05_tasks.md is agent-maintained — the agent updates it as work progresses, marking each task done and adding a one-line deviation note if implementation diverged. The plan files themselves become the status; Bridge doesn't track it externally.

How chat interacts with a plan

On the Refine tier, chat can read AND write plan files. When you scope a chat to a plan, the AI can propose edits that you see as inline diffs. You approve each diff, and the change writes through to the plan file.

See Chat for the chat-with-plan workflow.

Tier gating

Build Plan generation starts at the Discover tier. Discover includes one Build Plan a month: Start a venture works until it's used, then tells you when the next one is available (the 1st of the month, UTC). A plan that fails doesn't count. Build and above have no monthly limit, and the Refine tier adds plan-editing chat. Free tier can read plans you've already generated (when shared) but can't generate new ones.

See Tier comparison.

Next