The relationship engine your AI agent can actually run.
Relaytr finds the right people for a goal you describe in plain language, researches them with evidence — never invented facts — scores the opportunity, and recommends what to do next. Every recommendation clears jurisdiction, consent and policy before it's ever actionable, and nothing sends without your approval. Built to be operated by a person or an AI agent — the rules are the same either way.
Free to start — 250 contacts checked every month, no card required.
Not a scraper. Not a black-box AI SDR. Not another list your legal team can't sign off on.
32
UK, EU/EEA & US countries with jurisdiction-aware rules
0
Seat tax or list-size cap on stored contacts
100%
Of sends held behind an explicit approval step
10
Agent-operable tools, permission-checked like every other action
From a goal in plain language to a real, safe conversation
Seven steps, run by you or by an agent acting on your behalf — the rules don't change either way. Nothing downstream of step 5 is ever actionable until a contact clears it.
- 1
Describe your goal
“Find VPs of Engineering at Series B fintech startups in Germany.” Plain language in, structured search criteria out.
- 2
Find real people
Searches your connected provider (Apollo, ZoomInfo — bring your own account), deduplicated against everyone you already know about.
- 3
Research, evidence-backed
Every fact traces to a real, named source. Nothing is invented to make a message sound more personal than it is.
- 4
Score, and see why
A transparent, deterministic score — fit, evidence, relationship stage — not a model's confidence number you have to trust blind.
- 5
Clear the compliance pipeline
The 8-stage pipeline below: jurisdiction, consent, policy, throttling. Nothing downstream is actionable until this passes.
- 6
Approve, then engage
A human signs off. Automated where a real channel exists (email, SMS); a clear, tracked task where it doesn't — LinkedIn stays human, since there's no API for that to automate honestly.
- 7
Run it from an agent, safely
Same tools, exposed over MCP. An agent can search, score and prepare — but hits the identical permission and approval gates a person would.
Step 5, in full: one decision pipeline, eight stages, fixed order
This is the layer every relationship Relaytr helps you start has to clear first. It runs between a rep hitting “import list” and a message going out. Every approach produces exactly one immutable decision record — a verdict, a reason and a policy version — so when legal asks why a contact was reached, the answer already exists.
What the pipeline actually produces
Not a verdict in a log line — a full record. Real field names, real reason codes, two runs that stopped at different stages of the same pipeline above.
Trace fields differ by which stage stops a run — the consent-stage block carries no matched policy rule, because policy already passed this recipient through; the throttle-stage queue carries the throttle policy version that gated it instead. That variability is the schema, not a rendering gap. (Illustrative examples, generated from a synthetic recipient — not live customer data.)
Built for revenue teams, not marketing generalists
Not a faster sender. Not a compliance checkbox bolted onto someone else's tool. Autonomous relationship acquisition starts with knowing who you're allowed to reach — the combination below is what lets a RevOps or sales leader run UK/EU outreach without a standing legal review queue.
Compliance is a product feature
Not a paragraph in the terms. The reason a UK/EU revenue leader chooses this over outreach tools built for a market where none of these rules apply.
No provenance, no send
Every audience record retains source provenance and a lawful basis. Without both, an approach cannot be made.
No bypass path
Every approach passes the decision pipeline — not for admins, not for the API, not for a single test send.
Explainable by design
Every verdict is a single immutable decision record: reason, policy version, evidence. No black box.
Isolation, enforced in the database
Row-level security on every tenant-owned table, not just the service layer. Proven by test, not assumed.
Built to be handed to an agent, safely
Relaytr ships as an MCP server, not just a dashboard. Point Claude — or any MCP client — at it, and it can search, score and prepare next steps on your behalf, through the exact same permission checks a person hits in the product. That's not a policy in the docs — it's enforced the same way every other sensitive action here already is, and it's proven with a test.
No parallel authority
Every MCP tool call runs through the exact same permission check a person hits in the dashboard — never a separate, looser rule for machine callers.
No self-approval
An agent can prepare and recommend. It can never approve its own recommendation — that still takes a human with the right permission.
Every call classified and logged
Read, prepare or execute — spelled out per tool, not left to guesswork, and every call is written to an audit log before it returns a result.
Every channel a rep can legitimately reach a buyer through
Automated where a platform permits it. Generated for a human to send where it doesn't — cold social DMs, for instance, stay a human-only, flagged task, never an automated one, since automating them would breach those platforms' own terms. The distinction is always recorded, never blurred.
Start with knowing who's safe to reach — then let Relaytr find the rest.
Sign up, run a real compliance check against your own UK/EU contacts, and get a per-channel, per-recipient verdict — before a single rep, or agent, sends anything.
Free up to 250 contacts checked a month — no card required to start.