Try it now
Built for AI infra and dev tool companies

They dropped Pinecone on Tuesday.You should know by Wednesday.

Technography reads the public repos, issue threads, changelogs and job posts of your ICP. It then unearths which accounts are ready to talk NOW, and shows you the technical evidence behind it

Try it now Book a demo
Today: a ranked queue of accounts with a fresh signal and a real gap in fit, evidence beside each one
Issue #482 on a tool you trackEnterprise SSO support for agent runs
northwind-data/platformrequirements.txtcommit 9f2c1ab, 6 days ago
41langchain==0.2.14
42openai==1.40.0
43-pinecone-client==3.2.0
43+weaviate-client==4.6.1
44+datadog==0.49.1
45pydantic==2.8.2
Read as Vector DB migration, new monitoring Act now 3 lines of evidence. First seen 6 days ago. Stack signal.
Opportunity window card naming a gap in a company's tooling
40,000+Technologies tracked
1,000+Signals found every day
What we read

Different places a stackgives itself away.

Companies don't announce what they're switching to. They leave it in a repo, a thread, and a job post. We read all three, for every account in your market.

northwind.io/changelog

v2.9 moves vector storage to Weaviate

We've replaced our hosted index with Weaviate running in our own cluster. Query latency drops and we can finally keep customer embeddings inside our VPC. Monitoring now ships through Datadog.

Published 6 days ago
A stack signal. A vendor swap, announced by the company itself, with the date it happened.
Issue #482, opened by mchen

Enterprise SSO support for agent runs

We're evaluating vendors for our agent sandbox and SAML is a blocker. Is there a supported way to enforce SSO across shared runs, or do we need to build it ourselves?

openopened 3 days agoupdated today
A dev signal. A named engineer at the company, on a tool you track, asking the question you answer.
Careers, Northwind Data

Platform Engineer

You'll own the infrastructure our AI features run on.

  • Production Kubernetes and Terraform
  • Observability with Datadog
  • Vector search at scale
Posted 3 days ago2 similar roles open
A hiring signal. Open roles judged against your own product, not against AI in general.
How it works

A short list,in order.

Technography turns all of that reading into one queue: who to contact first, why now, and who to ask for.

Accounts: every account that matches your ICP, with signals, fit and relevant tech
Opportunity window and signal tabs for a watched company
Company sidebar with firmographics, infrastructure, buying committee and history
Today's queue of Act Now accounts, with evidence for the one selected
SafetyKit, buying committee
ChampionHead of Risk and Policy
InfluencerSenior Software Engineer
EnablerHead of GTM Strategy
How we justify it

Every claim showsits working.

This is the Fit & Intent view for one account, rebuilt from the product. Hover any outlined phrase to see whether it was observed or inferred, how confident we are, and the evidence behind it.

  1. Seen

    A public signal

    Three open roles for agent runtime, voice and conversation intelligence.

    Careers page, 3 days ago
  2. Evidenced

    Marked with its source

    Each fact is tagged observed and carries a confidence level, so you know what is a record and what is a reading.

    Observed, high confidence
  3. Inferred

    What it points to

    Roles plus a platform launch suggest the company is shipping agents to production.

    Inferred, medium confidence
  4. Acted on

    An angle to use

    The reason to call, in one sentence, with timing: a same-day cluster of signals.

    Fit strong, timing now
← Accounts

Halcyon Voice

halcyonvoice.aiReady
Write outreach
OverviewSignals 6StackCommittee 2
Angle
  • Reach out around Halcyon's active build-out of its agent-operable platform (new CLI, voice agent localization) and its hiring push for agent runtime, voice, and conversation-intelligence engineers. Frame Tracewell as the missing layer that keeps the full execution trace, tool use and voice pipeline integrations together in one trace, so their Python/TypeScript engineering teams can debug the production failures evals miss as they scale these new agent surfaces.
Timing triggerA same-day cluster of signals (CLI launch as an 'Agent-Operable Platform,' voice agent localization launch, and active agent-runtime, voice and conversation-intelligence hiring)
Why this is a fit
observedTarget's confirmed stack includes Python and TypeScript, the exact languages Tracewell is built for.
observedMultiple engineering roles for AI agent runtime, voice experience, conversation intelligence, and forward-deployed agent work indicate Halcyon is actively shipping production AI agents, the exact context where Tracewell's full-execution-trace debugging applies.
observedThe Halcyon CLI launch, framed as "A Developer Surface for an Agent-Operable Platform," shows Halcyon is building out developer-facing tooling around its agent platform, increasing the surface area needing production observability.
observedCompany size (450 employees), Series D funding, and US location match Tracewell's target company-size and geography bands, supporting reachability once use-case fit is established.
Gaps

OpenTelemetry and Tracewell are not detected in Halcyon's stack, suggesting no dedicated agent-level tracing or observability instrumentation is in place despite active agent development.

No common agent orchestration frameworks (LangChain, LlamaIndex, CrewAI, OpenAI Agents SDK) are detected, which is relevant since these are typical surfaces Tracewell instruments. Halcyon likely runs custom agent orchestration that still needs execution-trace visibility.

Why it is matched
PythonTypeScriptKubernetesPostgreSQL

Sample company and wording. Hover cards show how each claim is read; the evidence lines in them are illustrative.

Evidence first

The gap,tagged and sourced.

The opportunity card names what the company runs, what it doesn't, and why that matters to a seller. Click any screenshot on this page to see it larger.

Opportunity window for Halcyon Voice: runs LangChain, LangGraph and CrewAI, no LangSmith or Langfuse tracing tool detected
Company details: firmographics, infrastructure, buying committee and history

Beside the claim, the company itself: firmographics, its infrastructure, the buying committee, and its history with you.

Technology Graph

Look up any company,even one you don't have yet.

Type a domain. We resolve its stack, its infrastructure and its public GitHub org from public data, then check it against your ICP.

Try
A tool in your ICP profile
Anything else we found

Sample data, resolved from public data. Add a company to your ICP list to crawl its site and check it against your profile.

Account Search

Find accounts you don't have yet

Search by industry, headcount and location. DNS and certificate signals show per row as extra context.

Account Search with industry, headcount and location filters and a list of companies
Get set up

From your websiteto a first list.

Give us your website. We read it to learn what you sell, pre-fill your ICP, and estimate the list before you commit to anything.

This is a sample run. Nothing is sent anywhere. Your real ICP takes about a minute once you sign up.

  • 1Enter your websiteThat is all we ask for up front.
  • 2We read itThe radar works out what you sell and who buys it.
  • 3Confirm your ICPReview what we pre-filled, then see your estimated list.
Onboarding: enter your company website to pre-fill your ICP
$ read acme.com
  1. Reading your site
  2. Identifying your product
  3. Mapping likely buyers
  4. Estimating ideal company fit
ProductB2B SaaS analytics platform
Buyer stackSalesforceHubSpotSegment
Size51 to 500 employees
Funding stageSeed to Series B
~842companies match your ICP (estimate), plus a free sample.

Know first.

Add your first accounts and start watching today. It takes minutes to get set up.

Try it nowBook a demo