Can Drevon build a lead list?

Yes. Describe your ideal customer in a sentence, or point Drevon at a Sales Navigator search, and it returns a CSV where every row carries the reason the account matched and the source behind it.

Does Drevon do account research before a call?

Yes. Ask for a pre-call brief and Drevon reads the company’s history, funding, hiring, and what the people in the room have said publicly, with a link on every claim. Agencies run these in batches.

How does Drevon find buying signals?

It reads the public sources where real-time buying signals actually show up: Reddit threads, LinkedIn posts, job changes, funding announcements, review sites. Every signal comes with the page it was found on.

What jobs can a GTM AI agent run in Drevon?

The ones a GTM team actually delegates: build a lead list, account research, account health checks, an account relationship recap before a renewal, competitor analysis, an AE-to-CS handoff brief, buying committee mapping, pre-call briefs. Each job is a prompt, and every output cites its sources.

Do I need a ZoomInfo, Apollo, or Clay contract?

No. Drevon reads public pages and the tools you’re already logged into. If you run Clay or HubSpot, Drevon exports to them.

How is Drevon different from gtm.ai?

gtm.ai is ZoomInfo’s interface for AI agents: your agent queries their database, and you pay for a ZoomInfo contract plus credits. Drevon needs no data contract. It researches in your own browser with your own logins and cites the page it read.

What AI subscription do I need?

Any one of Claude Code, OpenAI Codex, GitHub Copilot, Google Gemini, or Antigravity. Drevon drives the plan you already pay for; there is no separate AI bill.

Drevon is a macOS desktop app that does evidence-backed prospect research for GTM engineers. From one or two prompts it finds high-intent customers and researches them in depth: LinkedIn activity, Reddit threads, blog posts, podcasts, years of company history, with every claim linked to its source. It runs AI agents (powered by Claude Code, OpenAI Codex, GitHub Copilot, Google Gemini, or Antigravity) directly in your own browser, without adding new subscriptions or sending data to the cloud.

Does Drevon connect to my CRM?

Where your CRM publishes an MCP server (Model Context Protocol), yes: add it in Settings and its data becomes part of the research, so the agent can check what your team already knows about an account before it goes looking. Drevon takes any MCP server, not a fixed list. For everything else, it exports structured files to the tools you already run.

How is Drevon different from Clay?

Clay is a cloud-based enrichment tool that uses API calls and third-party data providers. Drevon runs agents in your own browser, logged into your real accounts (LinkedIn, Sales Navigator, your CRM). This means Drevon can pull data Clay cannot reach (saved searches, private LinkedIn signals, personalized CRM context) and it charges no per-row credit fees.

How is Drevon different from Apollo?

Apollo is a contact database with outreach sequencing. Drevon is an agent runtime that can pull from Apollo, LinkedIn, your CRM, and any web source simultaneously. Drevon does not sell you data. It automates how you gather and use it.

Does Drevon require a subscription?

No. Drevon is free to download and runs on top of Claude Code, OpenAI Codex, GitHub Copilot, Google Gemini, or Antigravity. These are plans you likely already have. There are no per-row fees, no credit system, and no new contract.

Can Drevon agents access LinkedIn?

Yes. Because Drevon runs in your own browser with your existing login session, agents browse LinkedIn, Sales Navigator, and any other tool you are logged into, exactly as you would. No datacenter IPs, no rate-limit workarounds.

Is my data safe with Drevon?

All data stays on your machine. Drevon does not upload prospect lists, CRM exports, or agent outputs to any cloud server. It runs entirely locally, which means no third-party storage, no compliance risk, and no data leakage.

Can I run multiple AI agents at the same time?

Yes. Drevon supports parallel agent execution. You can run 10 agents simultaneously, each with its own browser context, memory, and instructions. Each agent works independently and saves its output to your local filesystem.

What does "evidence of intent" actually mean?

Every row Drevon hands back carries the thing that put it on the list: the Reddit thread and the quote, the LinkedIn post and the date, the funding announcement and the roles opened after it. You can click through to the source. That is the difference between a list of people who match your ICP on paper and a list of people who are looking right now.

How long does prospect research actually take?

Based on our user interviews: 15–20 genuinely personalized LinkedIn reach-outs is a full day of research for a rep with 8–10 years of experience. Drevon researches the same batch in 15–30 minutes, reading profiles, posts, podcasts, funding history, and company background, citing the source for every claim so you can verify before you send.

What GTM tools does Drevon work with?

Two ways in. Any tool that publishes an MCP server connects in Settings, which covers the stack most GTM teams run: crms (HubSpot, Salesforce, Pipedrive, Attio, Zoho CRM), enrichment (Clay, Apollo, Breeze (Clearbit), People Data Labs), outreach (Instantly, Smartlead, Lemlist, Outreach, Salesloft), analytics (PostHog, Amplitude, Google Analytics 4, Mixpanel), workspace & export (Notion, Slack, Google Sheets). Everything else takes the second route, the structured files (CSV, JSON, Markdown) every run writes to your filesystem.

All releases

v1.2.0releasemcpconnectorsintegrationsagents

3 min read

Drevon 1.2.0: connect any MCP server to your agents

Drevon TeamAugust 21, 2026

Drevon 1.2.0 lets your agents use the tools you already pay for. Connect Linear, Notion, Sentry, a Postgres database, or anything else that speaks MCP, and your agent can read from it and write to it while it works.

Connectors live in the chat box

Open the ＋ menu next to where you type and there is a new row: Connectors. Add a remote server by pasting its URL, or a local one by giving the command that starts it. Switch it on for an agent and that agent has the tools on its next session.

You do not have to know anything about how the server talks. Paste the URL and Drevon works out the rest, including the older style of connection that some servers still use. If a server will not work with the agent you picked, the row says so instead of quietly doing nothing.

Signing in actually works

Most useful servers want you to sign in first, and this is the part that tends to break. Drevon handles it: click Connect, approve access in your browser, and come back to a green dot. Your access renews itself in the background, and if it ever expires the row tells you and offers a single button to fix it.

Your tokens are kept in the macOS Keychain, handed to the agent only for as long as a session runs, and wiped the moment it ends. They never appear in your connector list or in any log.

All five agents, one list

Add a connector once and turn it on for whichever agents should have it. Claude Code, Codex, Copilot, Antigravity and Gemini each read configuration differently, and Drevon writes whatever each one expects. Servers you already set up by hand for those tools can be imported in one click from Settings › Connectors, which is also where you test, edit and remove them.

Answers stay answers

Ask a question and you get an answer. Previously, an agent that used a tool to look something up would sometimes decide it also owed you a file, and the answer got replaced by a note about saving one. Asking "when is my next meeting?" now just tells you. Ask for a spreadsheet and you still get a spreadsheet.

Files you can actually open

A file the agent creates shows up in the chat as something you click. If it is in the chat, it opens. Files that were only mentioned, or that moved after they were written, no longer leave you with a dead link, and generated pages now render the way they were designed instead of arriving unstyled.

Antigravity agents in particular were writing their work somewhere you could not see. Their files now land in the agent's own folder alongside everything else.

One more thing

Google has discontinued Gemini Code Assist for individuals, and the Gemini CLI may now refuse to run on affected accounts. Drevon says so on the Gemini option rather than letting a run fail without explanation, and points you to Antigravity, which is Google's replacement and works today.