The Agent-First CRM
Your contacts. Actually understood.
ProvenContacts is built agent-first: an AI agent runs the CRM, you don't. Talk to it the way you would a sharp ops hire. It creates records, enriches contacts, builds forms, and wires up automations, all from a single conversation.
Closed beta. We're onboarding teams in small batches.
The problem
Most CRMs are filing cabinets you maintain by hand.
Data entry, stale fields, forgotten follow-ups, dashboards no one reads. The tool that's supposed to save you time becomes another thing to keep up with. Bolting a chat box onto that filing cabinet does not fix it. ProvenContacts is built the other way around: the agent maintains the data, and you just ask for what you need.
What you get
A CRM that works the way you talk.
Agent-first CRUD
Create, update, enrich, and clean records by conversation. The agent has a tool for every operation the platform supports.
Enrichment built in
Find a person or company, disambiguate by name and employer, and fill in the gaps from public data, with the source cited.
Custom record types
Track vendors, sites, events, anything at all. Define a new type in one sentence and the agent builds the field catalog.
Forms & landing pages
Spin up a public capture form or a landing page from chat. Submissions land as records and fire your automations.
Automations & webhooks
When this happens, do that. In-app notifications, field updates, and signed webhooks to your own systems.
Open API & MCP
A REST API at /api/v1 with bearer tokens, plus an MCP connector so Claude Desktop and Cursor can drive your CRM.
How it works
Up and running in minutes.
Connect your data
Import a CSV or paste a list. The agent maps columns and dedupes as it goes.
Talk to your CRM
Ask for what you need in plain language. The agent runs the operations and shows its work.
Let it run
Set automations once. Capture leads with forms. Push events to your stack with webhooks.
Questions
Good to know.
- Is this just a chatbot bolted onto a CRM?
- No. Every operation the platform supports exists as a tool the agent can call: creating and updating records, enrichment, custom record types, forms, automations, member admin. The chat is how you run the CRM, not a help widget stapled to the side.
- Can I still use a normal UI?
- Yes. Every capability has a familiar screen behind it: lists, boards, record pages, dashboards, settings. The agent is faster for most things, but nothing is chat-only, and you can always click.
- How does enrichment work?
- Run it from a record, a list row, a bulk selection, or just ask the agent; captured leads are enriched automatically in the background. Each run searches public sources (LinkedIn, the company's website, the open web), confirms the match by name and employer before anything is written, then shows you a preview: every proposed field with its source and a confidence score. By default it only fills empty fields; overwriting your existing values is an explicit opt-in, and if the match can't be confirmed it writes nothing rather than guessing.
- What about my data, and which AI models do you use?
- The agent runs on Claude, Anthropic's frontier models. We reach every model through its API under business terms that prohibit training on your data, so nothing in your workspace is ever used to improve someone else's model. Your records stay in your workspace; the agent only sends a model the context it needs for the task in front of it.
- Can other tools connect to it?
- Yes. There's a REST API with bearer tokens, outbound signed webhooks, and an MCP connector so MCP-aware clients can operate your CRM directly.
Stop maintaining your CRM. Start talking to it.
Join the waitlist for the closed beta and put the agent to work on your pipeline.