A Build Graph converts your research output into a concrete build plan. You select the features you want to build, and ManticScore generates a recursive tree of implementable tasks — each with a name, description, technical spec, acceptance criteria, and a risk rating. You can expand any node deeper, get an AI effort and confidence assessment, update task status as you work, and kick off a Forge run to generate code directly from a node’s spec. The graph grows with your project and stays in sync with your work.Documentation Index
Fetch the complete documentation index at: https://docs.manticscore.com/llms.txt
Use this file to discover all available pages before exploring further.
What a node contains
Every node in the tree is self-contained and actionable:green (straightforward), yellow (some unknowns), red (high uncertainty or external dependencies).
Node status tracks where the AI is in expanding this node: collapsed, expanding, expanded, leaf, or implementing.
User status tracks your progress: pending, in_progress, done, or skipped.
Creating a build graph
Select features
Choose which features from your research you want to include. Each feature needs an
id, label, and source. You can also pass auto_select: true to let ManticScore choose based on your research’s build readiness data.Submit the request
Send
POST /build-graphs. The graph is created and generation begins in the background — you receive a graph_id immediately.Features to include in the graph. Each item:
{"id": "string", "label": "string", "source": "string"}.Your product idea, up to 5,000 characters.
UUID of an existing project. Pass
null to leave unattached.When
true, ManticScore reads your research’s build readiness data and selects features automatically.Build graph events
The stream follows the NDJSON v1 protocol and supports cursor-based resume.| Event | Data |
|---|---|
stream_start | Stream confirmed open |
graph_created | Graph row created; includes graph_id and project_id |
stage | LLM stage started or completed |
node_created | A new node added to the tree |
graph_ready | Initial generation complete; includes root count and total duration |
expand_start | A node expansion has begun |
expand_done | Expansion finished; includes child count |
error | Non-retryable failure |
done | Stream closed |
node_created event includes:
Working with nodes
Once the graph is ready, you can interact with individual nodes.Expand a node
Nodes start collapsed. Expanding a node generates its children — the sub-tasks needed to complete it. Expansion runs in the background and emits events on the graph stream.Assess a node
Get an AI-generated risk and effort assessment for a node:curl
confidence (low, medium, or high) and effort (small, medium, large, or very_large), along with a list of unknowns the AI flagged.
Update user status
Mark a node as in progress, done, or skipped as you work:curl
name and description if you want to adjust the AI-generated text.
Implement a node with Forge
Turn a node directly into a pull request. ManticScore passes the node’s spec to Forge, which plans and writes the code autonomously.curl
GET /forge/runs/{run_id}/events.
Create a ticket
Export a node as a Linear or Jira issue, using your connected defaults:curl
Export to Notion
Export a node as a structured Notion page:curl
Fetching a complete graph
curl
nodes field.
Limits and credits
| Value | |
|---|---|
| Credit cost | 1 credit per build graph |
| Max idea length | 5,000 characters |
If the latest completed feature research for your project exists, ManticScore automatically injects it into the build graph generation prompt — no extra configuration needed.