AgentInfra Radar
Research methodology

How AgentInfra Radar turns public AI infrastructure signals into usable leads

AgentInfra Radar is built from public-source research, structured filtering, and manual verification. The goal is not to promise customers; it is to help GTM and research teams find relevant projects worth reviewing.

Public-source inputs
We review public material that buyers and researchers can inspect themselves.
  • GitHub repositories and organizations
  • MCP server directories, docs, and package pages
  • Official product sites, changelogs, launch posts, and blogs
  • Public community, integration, hiring, and partnership signals
Screening logic
A project must be relevant to the AI infrastructure buying map before it enters a lead pack.
  • Clear relationship to MCP, agents, LLM apps, evals, gateways, observability, security, or deployment
  • Enough public context to explain why the lead matters
  • A plausible need for a vendor, consultant, investor, or partner
  • Duplicate and low-context entries are removed or held for review
leadScore
leadScore is a practical prioritization aid, not a conversion forecast.
  • Category fit and buyer relevance
  • Freshness and clarity of public signals
  • Evidence quality and source confidence
  • Actionability of the suggested contact angle
growthSignal
growthSignal summarizes the public reason a project is worth watching now.
  • Recent releases, docs, integrations, or launch activity
  • GitHub activity, community attention, or adoption references
  • Hiring, partnership, customer, or ecosystem expansion signals
  • Signals are reviewed as context, not as guaranteed intent
verificationStatus
verificationStatus helps teams know whether a record is export-ready or still under review.
  • Public-source verified: core fields are backed by public pages
  • Needs manual review: one or more fields need human confirmation
  • Monitor only: relevant project, but not ready for outreach
Why manual review stays in the loop
AI infrastructure categories shift quickly. Manual review prevents generic, stale, or misleading outreach.
  • Confirm the project is still active and correctly categorized
  • Check whether the contact angle is specific and respectful
  • Avoid using private, guessed, or scraped personal contact data
  • Flag leads that should be watched rather than contacted
Next step

Want to inspect the format first?

Review the beta sample pack before asking for a larger list.

View sample pack