Perch is an AI built on everything Texas legislators say on the record — committee testimony, official statements, press releases. Run any member, bill, or issue and get where they actually stand, cited to the statement that proves it.
Walk in knowing what they'll say.
Request Access →A member, a committee, a strategy — the same engine reads all of it and answers from the record. Try one:
Thousands of attributed statements resolve into one answer — every line traceable to the statement it came from.
Primary Sources Only
Each behavioral read is a named technique from political-science and NLP research — not a vibe — applied to the committee record and reported with its sample size.
1,536-d embedding; each member’s centroid is scored by cosine similarity against all 181 peers to find who they sound most like.inquiry, position-taking, or challenge — and by who it’s aimed at.Every measure carries its n. Below minimum sample size, we leave it blank rather than guess. Embeddings: OpenAI text-embedding-3-small (1,536-d). Refreshed weekly; every score links back to the cited statements behind it.
If your job is knowing where a member stands before you walk in, Perch is the prep.
Already have access? Sign in →
A Perch search can reveal who you’re working for. So it’s built to keep that to yourself — encrypted, private, and never used to train a model. Privacy Policy →
Your searches, briefs, and history aren’t visible to other subscribers, to the members you research, or to anyone at Perch. Yours alone.
Encrypted in transit and at rest. Production access is locked down, and billing runs through Stripe — we never touch your card number.
Your queries and briefs are never used to train AI models — not ours, not anyone’s. Your strategy doesn’t become someone else’s feature.
Every claim links to the statement it came from. AI can err — the [s1] citation is how you catch it, in one click against the record.