Natural language search
Type "fintech founders" or "accountants in Toronto". Ten free sources, merged and de-duplicated.
The open-source lead workspace. OpenLeads searches free public sources, finds who to email, labels every address as found or guessed, then writes, sends, and repeats it on a schedule.
Real counts from the nightly benchmark, run 7 Oct 2026. Ten leads requested per search. See every search
Live benchmark
Every night, OpenLeads runs 12 real searches against live sources and commits the results to the repo.
12/12
searches returned leads
84%
of requested leads delivered
32%
of emails backed by evidence
0%
junk names in the results
Workflows
Finding, verifying, writing, and sending usually take four paid tools. In OpenLeads they are nodes on one canvas. Build the flow once, then let it run every weekday at nine.
The workspace
Run openleads web and you get a full dashboard: search, a CRM, sequences, analytics, and visual workflows. No account, no seat pricing, no data leaving your laptop.
Type "fintech founders" or "accountants in Toronto". Ten free sources, merged and de-duplicated.
Paste a list of companies or names. Every row comes back with an email and a label you can trust.
Recipes drawn as a node canvas: trigger, find, verify, write, send, export. Run once or on a schedule.
Every lead tracked from found to replied, with statuses and suppression lists in one SQLite file.
Follow-ups that stop the moment someone replies, bounces, or unsubscribes.
Reply rate, bounce rate, and results by source and tier, computed from your own send log.
Hear about it only when new leads match a niche you care about, delivered to a file or a webhook.
Let Claude, Cursor, or any MCP client find and enrich leads for you, with the same labels.
SPF, DKIM, and DMARC graded before your first send. Warmup caps and throttling are on by default.
Sends through and exports to
Deliverability
Most free finders rely on one SMTP probe that home ISPs block. When it fails, every address quietly becomes a guess. OpenLeads cross-checks seven signals instead.
Most of them need no port 25 at all.
Every address gets a 0 to 100 score and lands in a tier.
You never pick one. Your search is routed to the sources that fit, then merged and de-duplicated.
Compare
| OpenLeads | Apollo and Hunter, free | |
|---|---|---|
| Cost | $0, with no credit limits | Credits, then paid |
| API key | None | Required |
| Labels | Found, pattern, or guessed on every address | One confidence score |
| Enrich your own list | Paste a CSV, get each email labeled | Paid feature |
| Writes and sends | Drafts, warmup, throttling, suppression | Upsell |
| Workflows | Visual recipes on a schedule | Paid plans |
| Where it runs | Your machine | Their cloud |
| Source code | Open source, AGPL-3.0 | Closed |
Leads, drafts, mailbox credentials, and your CRM live in one SQLite file under ~/.openleads. The engine only talks to public sources and your mail server.
Pricing
No credits, no seats, no trial that runs out. The whole product is open source. Companies that cannot ship AGPL code can buy a commercial license.
$0forever
The full product, self-hosted on your machine.
1 staron GitHub
Same product. You help the next person find it.
Custom
For teams embedding OpenLeads in closed-source products.
The engine only uses free public sources and your own mailbox, so there is nothing to meter. The code is AGPL-3.0. If you want to build it into a closed-source product, there is a commercial license.
Every address is labeled: found (published somewhere), pattern (matches the company's known format), or guessed. The nightly benchmark publishes how many of each we return, so you can check before you trust it.
Only safe-tier addresses are sent by default. Sends are throttled and capped during warmup, every email has one-click unsubscribe, and the preflight grades SPF, DKIM, and DMARC before the first send. Use a secondary domain for cold outreach.
It depends on where you and your recipients are. OpenLeads ships suppression lists and one-click unsubscribe headers, and the responsible use guide covers CAN-SPAM, GDPR, and CASL. It is built for relevant, low-volume outreach, not bulk spam.
Recipes are workflows: a trigger, a search, a verify step that branches on the result, then write, send, and export nodes. The dashboard draws them as a canvas you can edit and run. You can also push leads into n8n or Zapier with the webhook export.
No. Install it with one command, or run npx openleads and type what you want in plain English. Python only matters if you want to add your own source.
Python 3.8 or newer. The npx wrapper runs the same CLI.
# the engine and the local app
pip install "openleads[all]"
# describe who you want
openleads find "marketing agencies in Miami"
# or open the local dashboard
openleads web# no install, wraps the Python CLI
npx openleads find "rust developers in Berlin"
# chat mode: just type what you want
npx openleads
# add your own source
cp lawyers.py ~/.openleads/sources/# add OpenLeads to Claude Code as an MCP server
claude mcp add openleads -- openleads mcp
# then ask in plain English
find 20 accountants in Toronto and label their emails