Channels
AI CRM
By team
Alternatives
Learn
Guides
Open source · self-hosted
AgentSDR runs your email sequences, LinkedIn campaigns and WhatsApp calls, and an AI CRM that reads every reply and drafts the answer. On your server, with your own model key.
The real app with sample data. No sign-up.
Channels
Email, LinkedIn and WhatsApp run side by side, and every reply lands in one AI CRM. Pick a channel, or one of its capabilities, to see it work.
Sequences from your own Google Workspace mailboxes.
Emails sent
2,999
vs 2,690 previous period
Leads contacted
1,545
vs 1,402 previous period
Replied
118
of 1,545 contacted
Reply rate
7.6%
vs 6.9% previous period
Bounced
21
vs 34 previous period
Unsubscribed
9
vs 12 previous period
Emails sent each day: first touches and follow-ups.
First touch
Follow-ups
Right now, regardless of the date range.
Sent today
108 of 150 sent today
Inbound email replies each day.
Replies
Opens and clicks aren't recorded, so there are no open or click rates here. Reply rate is the engagement metric to watch.
Sends are for this period; leads, replies and bounces are all time. Sorted by sends.
| Campaign | Leads | Contacted | Sent | Replied | Reply rate | Bounced |
|---|---|---|---|---|---|---|
| 1,240 | 812 | 1,388 | 46 | 5.7% | 7 | |
Agencies · Founders (EU) Active | 860 | 544 | 902 | 31 | 5.7% | 5 |
| 420 | 402 | 611 | 22 | 5.5% | 4 | |
Webinar attendees · Sept Paused | 310 | 196 | 244 | 12 | 6.1% | 3 |
Dev-tools CTOs · pilot Completed | 180 | 180 | 118 | 7 | 3.9% | 2 |
Day to day
Mailboxes pacing themselves, LinkedIn accounts inside their limits, calls recorded as they happen, and every reply answered before you open the inbox.
Sent today
108 of 150 across 5 mailboxes
Multi-step email from your own Google Workspace mailboxes: any CSV column as a {{merge field}}, spin text, a daily cap and sending window per mailbox, one-click unsubscribe and automatic bounce suppression.
Campaigns across several accounts through Unipile: an invite, an accept message and three follow-ups. 30 invites a day on premium, 5 on free, with a randomised gap between each.
Dial from AgentSDR: a Chrome extension places the call inside WhatsApp Web, records both sides, and your own model writes the transcript. Messages sync too, with a warm-up for new numbers.
Hannah Weiss· 14m
Could you do Thursday afternoon? I'll bring our ops lead.
Thursday works — I've held 3:00pm CET and sent the invite. I'll bring the reply-rate numbers from teams your size.
The CRM classifies each reply — interested, customer, not interested, other — and drafts the answer from your knowledge base. It only moves a lead forward on its own; anything else waits for you.
Data & AI
One database of people and companies feeds every channel, and every AI step runs on your own model key.
| Name | Job title | Source |
|---|---|---|
| Hannah WeissLumen Freight | Head of Growth | CSV import |
| Rafael CostaBrightloop | VP Sales | LinkedIn search |
| Aiko TanakaKestrel Health | Growth Lead | Table |
| Samir HaddadParcelly | COO | CSV import |
People and companies shared by every channel. Import CSV or XLSX, add your own columns; duplicates match on email or LinkedIn.
Columns that call an API, run a formula, ask your model or pull from Apollo — then become a campaign.
Bring your own model
Every AI call runs on your OpenRouter key, pinned to the provider you chose — no silent fallbacks.
Your data stays yours
Leads, conversations and call recordings live in your Postgres and your own storage bucket.
Guardrails on every channel
Daily caps, sending windows, warm-up for new numbers, and Do Not Contact honoured everywhere.
Drafts, not auto-sends
The AI proposes and a person approves. Low-confidence classifications wait for review.
Grounded in what you know
Drafts draw on your knowledge base: pricing, FAQs, case studies and objections.
Keyboard-first
⌘K jumps anywhere; J and K move through an inbox, Enter opens.
Self-host
Free and open source. No seats and no per-contact fees — you pay for your server and your own AI usage.
# 1 · Clone$ git clone https://github.com/Kandid-ai/AgentSDR.git && cd AgentSDR# 2 · Configure$ cp .env.example .env POSTGRES_PASSWORD=… BETTER_AUTH_SECRET=… BETTER_AUTH_URL=… INTEGRATION_CREDENTIALS_KEY=…# each secret: openssl rand -hex 32# 3 · Postgres, schema, app and scheduler$ docker compose up -d✓ Ready — http://localhost:3000What you connect
Your own accounts, so the data and the sending reputation stay yours.
Blog
Sales StrategyOutreach vs Salesloft: Finding the Right Engagement HubComparing Outreach vs Salesloft side by side so your team can pick the best sales cadence software for pipeline, pricing, and workflow fit.
Sales StrategyBeyond Enterprise: Alternatives to Legacy Sales SoftwareSkip bloated enterprise platforms. Compare modern B2B sales execution tools built for SDRs, founders, and revenue ops who need speed and flexibility.
Sales StrategyPrivate Customer Database: Secure Self-Managed CRM OptionsCompare private customer database options you control yourself, with practical guidance on secure, self-managed CRM alternatives to mainstream platforms.F.A.Q
Yes. It is open source and you host it yourself, so there are no seats, tiers or per-contact fees. What you pay for is your server, the accounts you connect (Google Workspace, Unipile) and your own AI usage on OpenRouter.
Google Workspace mailboxes, connected through a service account with domain-wide delegation. Each mailbox has its own daily limit (30 by default), sending window and signature, and several campaigns can share the same pool. Other providers aren't supported yet.
No tool can promise that, but AgentSDR stays well inside LinkedIn's limits: 30 invites a day on premium accounts and 5 on free, a random 30–60 second gap between invites, sending only inside each account's working hours, and an account that hits LinkedIn's own limit pauses for the day.
You link your number through Unipile and install the AgentSDR Call Recorder Chrome extension. When you call a lead from AgentSDR, the extension dials inside WhatsApp Web and records both sides; the recording goes to your storage bucket and your chosen model transcribes it. It relies on WhatsApp Web's English interface.
Any model on OpenRouter, with your own key. Each call is pinned to the provider you picked, with fallbacks turned off, so your data only goes where you chose. Classification, reply drafts, transcription and AI table columns all run this way.
No. Drafts wait in Action required until someone sends them, as written or edited. The AI can move a lead forward in the pipeline when it is confident; a backward move, a low-confidence call, and every new Customer are held for a person.
In the Postgres database you run, and call recordings in your own Cloudflare R2 bucket, reached only through short-lived signed links. Provider keys are encrypted at rest with AES-256-GCM. There is no hosted AgentSDR service in the middle.
Yes. Everyone signs in with their own account, and you invite teammates into your organization as owners, admins or members. One deployment can also hold several organizations, each with its own leads, inboxes and connected accounts, fully separate from the others.
A machine that runs Docker, and PostgreSQL 16 or newer (the Compose file brings its own). Copy the example environment file, set the database URL, an auth secret and an encryption key, then connect each channel from inside the app. Start with the quick start on GitHub.
That's the point of owning the code. It is a TypeScript Next.js app on Postgres, so adding a column type, an enrichment provider or your own channel is ordinary application work — and issues and pull requests are welcome on GitHub.
More questions? open an issue or read the code
Get started
Clone the repo, point it at Postgres, connect the channels you sell on. Your leads, your conversations and your model key never leave your infrastructure.
Analytics
How conversations turn into meetings and customers, and which channels bring them in.
Sample data
Replies
512
from 386 people
Positive
104
27% of people who replied
Meetings
37
vs 29 previous period
Customers
6
vs 4 previous period
Median first response
38m
vs 1h 11m previous period
Inbound replies each day, by the category they were classified into.
Interested
Customer
Not interested
Other
Unclassified
Right now, regardless of the date range.
Open records by category, right now, and how far along the funnel they are.
Records moved into a new stage, by the category they moved into.
Every record by workflow state, right now.
262 drafts generated in this period.
Drafts sent
214
AI drafts that went out
Classified
498
Replies the AI labelled
Override rate
6.4%
Labels a person changed
71% of sent drafts went out unedited (151 of 214)
151 62%
63 26%
29 12%
Where these conversations came from. Open a channel for its detail.
How each channel performed in this period. Select a row to open it.
| Channel | Reached | Replied | Reply rate | Positive | Open |
|---|---|---|---|---|---|
| 2,874 | 118 | 4.1% | 29 | ||
| 1,902 | 171 | 9% | 47 | ||
| 542 | 97 | 18% | 28 |
From first touch to customer.
Counts for this period — not one cohort followed through.
