Skip to content

icp.md

Latest commit

 

History

History
66 lines (48 loc) · 3.02 KB

icp.md

File metadata and controls

66 lines (48 loc) · 3.02 KB

ICP — Threadplane

Operational doc. Edited by humans as we learn. The durable two-track ICP summary lives in gtm.md §3.

Developer track

Buyer: Angular engineer (senior / staff) actively building or evaluating an agent feature for a production Angular app.

Buying signals

  • Lands on a comparison page (/compare/*) from organic search or referral.
  • Copies the install command (hero_install or compare_*_install CTA).
  • Starts a cockpit recipe with a stack hint matching their backend (LangGraph, AG-UI, custom).
  • Returns to the same comparison page within 7 days.
  • Hits the six-signal activation funnel and drops off at "thread_persisted" or "interrupt_handled" — those drop-offs are the most actionable.

These are the signals to instrument first in Spec 1. Baseline rates between them populate this section quantitatively once developer-funnel ships.

Disqualifiers

  • Greenfield app with no Angular constraint — Threadplane is for teams that must stay on Angular.
  • Hobby/learning project — we welcome them, but they aren't the ICP for prioritization.

Where they hang out

  • Angular Reddit, ng-conf community, Angular Discord/Slack equivalents, LangChain forum, AG-UI / A2UI communities, dev.to/medium.

What they need to see in 30 seconds

  • "Threadplane — Production-ready chat, threads, and generative UI for AI agents."
  • LangGraph + AG-UI · Durable threads · Interrupts · Subagents · Planning + memory · json-render + A2UI
  • One install command.
  • A working demo they can clone.

Enterprise track

Buyer: Architect or engineering lead at an Angular-heavy company (SaaS, internal tools, fintech, healthcare, insurance, manufacturing, compliance-heavy workflows).

Buying signals

  • Already using LangGraph or evaluating an agent framework.
  • React rewrite is politically or technically impossible.
  • Has a defined design system and cannot ship arbitrary JSON-rendered components.
  • 50+ engineers in the Angular org.

Disqualifiers

  • React-only frontend stack.
  • "We'll figure out the UI later" — no committed agent project.
  • Pure dev-tools company without an end-user-facing surface.

Qualified-lead v1 definition (server-side)

  • Non-personal email_domain (rules out gmail/outlook/yahoo).
  • Non-empty company field.
  • track=enterprise (the surface they came from).

Tighten in Spec 7 if signal proves noisy.

Vertical priorities (initial)

  1. Fintech / insurance / compliance — strongest fit for human-approval flows.
  2. Healthcare — adjacent compliance + design-system rigor.
  3. Internal-tools-heavy SaaS — high agent-affinity, low channel cost.
  4. Manufacturing — interesting fit for tool-call lifecycle UX.

What we do not yet know

  • Conversion rates between funnel stages (Spec 1 will produce baselines).
  • Geographic distribution of qualified leads.
  • Time from first $pageview to lead_qualified for the enterprise track.
  • Average time-to-first-message for developer track (cockpit instrumentation needed).

This document is updated when these become measurable.