AgentSDR is an open-source AI SDR (Sales Development Representative) tool designed to consolidate the multiple tools commonly used in the outbound sales process into a single self-hosted workspace. Traditional outbound typically requires stitching together a prospect database, data enrichment tools, email sequencing tools, LinkedIn tools, a dialer, and a CRM, each with its own seats, its own copy of prospects, and a different understanding of reply status. AgentSDR replaces these tools with one workspace, where each person and company has only one record, and replies on LinkedIn appear in the same queue and pipeline as email replies.
Key capabilities:
- Unified records across three channels: supports sending emails from your own mailbox, LinkedIn invitations and messages, and WhatsApp chats and calls, with all channels sharing the same prospect record.
- Built-in data enrichment: tables can run 15 data providers (such as Apollo, Hunter, Lusha, etc.), support AI columns and HTTP calls, and process row by row.
- Email sequences: sent through Google Workspace mailboxes, with support for merge fields, sending time windows, and daily limits.
- LinkedIn outreach: send connection requests with notes, follow-up messages after acceptance, and comply with conservative daily limits.
- WhatsApp chats and recorded calls: supports message campaigns, make calls from AgentSDR, and recordings are saved to the user's own storage bucket and transcribed.
- AI CRM: the AI reads every reply, classifies it as interested, meeting requested, not considering for now, out of office, etc., and displays the reason and a drafted reply.
- One-click approval to send: drafts are editable and sent through the channel the prospect replied on, and are never sent automatically.
- Automatically updated pipeline: each prospect is in its current stage, and follow-up sequences are scheduled automatically.
- Analytics: view replies, positive replies, meetings, and customer conversions by channel and over time.
How to use: AgentSDR requires a machine with Docker installed. Download one file, set a few keys, run docker compose up -d, and open http://localhost:3000，设置页面会创建管理员账户和组织。然后在设置中连接服务。对于服务器部署，需先替换 The # CHANGEME values in docker-compose.yml (or the adjacent .env) can be generated with openssl rand -hex 32. Running from source is also supported (requires Bun 1.2+ and PostgreSQL 16+), as well as deploying to a VPS with Dokploy and moving into production.
Use cases: Suitable for sales teams, startups, or developers who need to keep control of their outbound data and processes. Users own their infrastructure, and prospects, conversations, and recordings never leave servers they control. Users connect their own Google Workspace, Unipile, storage, and any AI model on OpenRouter, with no per-seat or per-contact fees—only server and provider costs.
Limitations and caveats:
- LinkedIn's terms do not allow automation, and accounts may be restricted; WhatsApp bans numbers that send unsolicited bulk messages; cold email is regulated by laws such as CAN-SPAM, GDPR, and CASL; many places require consent before recording. Sending limits, unsubscribe links, and Do Not Contact reduce risk but cannot make any use compliant.
- Microsoft 365 mailboxes are currently not supported.
- WhatsApp call recording depends on the English interface of WhatsApp Web.
- AI drafts (including follow-up sequence steps) wait for manual sending and are not sent automatically.
- The project uses the MIT license and may be used commercially, modified, and built upon.
