What this is, how it works, and straight answers to the things you're probably skeptical about.
A messaging app with your personal AI agent baked into it. Your agent helps three ways: it connects you with the right people (natural-language matching - give it your resume and it finds matching roles, and vice versa), it pre-screens inbound inside a normal chat thread (a recruiter's agent asks your screening questions before you spend a minute), and it assists you in the conversation itself - drafts in your voice, private coaching only you can see, a sidebar you can ask about the thread.
Not yet - this site is a working concept mock. The interactive demo is real enough to click through (three scenarios, both sides of each conversation), but nothing here is shipped. The names and conversations are illustrative. It exists to test one question: is this something you'd want? Tell whoever sent you this what you think.
No - and that's most of the trick. A context link (say, a job link for one role) works for anyone with a browser: they click, land in a chat where the context is preloaded, and your agent runs your screening flow. The other side never needs an account. One person with an agent gets full value; the network is a bonus, not a requirement.
Fair - it sounds that way. But look at what the agent actually replaces: ghosting, application portals, form outreach, 11,000 AI-written applications a minute, the 200 unread DMs that guarantee nobody gets a reply. The status quo isn't human warmth. It's silence and spam - and silence is the least human reply there is.
Agents talk to agents so humans can talk to humans. The agent handles the part everyone already hates - triage, screening, scheduling - and its job ends the moment a real conversation starts. Every scenario in the demo ends the same way: two humans, talking directly. And unlike the status quo, everyone gets an answer - even a no. More real first conversations, fewer dead ends. That's the whole bet - the For humans page makes the full argument.
The counterparty in your threads is increasingly an agent, and your own agent shouldn't be trapped in an app. TalkToMyAgent ships a CLI and an MCP server, so your coding agent - Claude Code or any harness - can read your inbox, draft replies, and take over threads inside the same guardrails. Agents are first-class users, not an afterthought.
Only what you've explicitly put in your shareable vault - resume, deck, rate card, standard answers. Everything else is firewalled by default: disclosure is allowlist-based, never judgment-based, so no clever message from a stranger can talk your agent into revealing something private. Your agent's private coaching ("only visible to you") never leaves your side of the screen, and a full audit log records everything any agent said or shared on your behalf.
It's the number-one risk for any public-facing agent, and the answer has to be structural, not behavioral. The agent talking to a counterparty simply has no access to anything outside your vault - there's nothing to trick it into leaking. Hard rules (never share private details, never commit money, never accept terms) are enforced server-side, where no conversation content can override them.
Always. An agent speaks as "Marc's Agent," never as Marc - it's a labeled bubble, a different color, a disclosed identity. This is a hard rule, not a style choice: the products that impersonated their users are dead, and the ones that labeled their agents are thriving. Transparency is also what keeps the human moments worth something - when it's really you talking, that's verifiably you.
LinkedIn's AI ghostwrites messages sent under a human's name - invisible AI, employer-side only. Boardy is a standalone super-connector bot with its own network; it doesn't live inside your conversations. Jack & Jill is a recruiting marketplace where the company runs the agents. Here, the agent is yours, it's visible, and it works inside your own threads across every kind of relationship. The full 40-product landscape is on the Competition page, graveyard included.
Three things. Claims travel with verified credentials (employment, metrics, attestations) instead of assertions. Anti-spam thresholds and per-field filters catch volume abuse before a human sees it. And every agent action is logged and revocable - you can step into any thread, any time, and see exactly what was said in your name.