Overview
What it is
Soorce AI helps recruitment teams find and contact candidates on LinkedIn. A recruiter pastes a job description. An LLM turns it into LinkedIn search filters and a short list of weighted screening requirements, and every profile found is scored against each requirement and ranked. The recruiter then runs multi-step outreach (invites, messages, InMails, follow-ups) from their own LinkedIn account. It was built inside Seleqt from July 2024, rebranded as Soorce AI in April 2025, and I have been its CTO (part-time) since the Soorce AI partnership started in October 2025. More than 100 organizations use it.
I wrote its first commit and built the Django backend, the React frontend, the async screening pipeline, the outreach scheduler, the shared inbox, a public API, billing, and an MCP server with its own OAuth authorization server.
Outreach has contacted 150K+ candidates at a 23% average response rate. The screening pipeline handles 11K+ profiles per day, and has screened 1.5M+ across 6,000+ roles.
Why it’s interesting
- An LLM pipeline screens and ranks thousands of LinkedIn profiles per vacancy against weighted requirements, more than 11,000 a day in production.
- Outreach runs itself: multi-step LinkedIn sequences that respect sending windows, time zones and daily limits per account.
- Guardrails stop double contact: blacklists, candidates already contacted from another campaign and new replies are checked before every send.
- Recruiters can use Soorce AI from Claude, ChatGPT and other AI assistants through an MCP server with secure sign-in.
- 100+
- organizations using it
- 150K+
- candidates contacted, 23% average response
- 11K+
- profiles screened by LLMs per day
- 1.5M+
- profiles screened across 6,000+ roles
company figures, Nov 2026
company figures
company figures
company figures
Try it
Screen a shortlist like the pipeline does
Requirements extracted from a vacancy are weighted and every candidate is scored against them; change the weights and watch the ranking change.
Screens
A closer look
Start from a vacancy. A recruiter pastes a job description, uploads a CSV, or points at a profile to find similar people.
1 / 10Tap a screenshot to enlarge it.
Features
Everything it does
Campaign creation
- Four ways in — Pasted or uploaded job description, CSV import with column mapping, or similar-candidate search from a profile.
- LLM search filters — Title variants, location with radius, language and ideal experience, plus manual filters and presets.
- AI signals — Job hopper strictness, freelancer, startup or corporate background and overqualified, as include or exclude filters.
- Weighted requirements — Generated from the vacancy, editable, each weighted 1x to 5x.
Screening and candidates
- Live async screening — Progress polling and early reveal at 90% so recruiters can start reviewing.
- Ranked review — List and table views, match percentage, per-requirement score and reason, notes and labels.
- Bulk actions — Shortlist, archive, move to another vacancy, duplicate management, blacklist by company or profile.
- Talent insights — Experience and location distribution, an AI salary estimate and a rubric-based feasibility score.
Outreach
- Workflow builder — Invite, InMail, message, profile visit and like-post steps, with branching conditions.
- Message templates — Variables, attachments, AI drafts and per-candidate previews; workflow templates per workspace.
- Sending limits — Per-weekday windows, time zone, daily caps per action type, skip people contacted elsewhere.
- Run control — Start, pause, resume, reschedule failed actions, pause or stop a single candidate.
Inbox and analytics
- Shared inbox — Search, unread, flags, notes and replies, with executed-action history per conversation.
- Auto labels — An LLM labels conversations such as interest, question or no interest.
- Campaign analytics — Invites, acceptance, replies, intent funnel, PDF export.
- Team dashboard — Workspace totals and CSV export.
Platform
- Sender accounts — Connect LinkedIn by hosted link, with tier-based limits and reconnect sync.
- Teams and billing — Workspaces, roles, blacklist, API keys, Stripe billing and onboarding tasks.
- Public API — REST API and Zapier triggers, plus market and talent intelligence endpoints.
- English and Dutch — Interface translated, and LLM output follows the user's language.
AI assistants and admin
- MCP server — Eight read-only tools: workspace, vacancies, candidates, replies, campaign steps and templates.
- OAuth consent — A consent page and a connected-apps list where members can disconnect assistants.
- Chrome extension — Described from its manifest and README, not run: links a LinkedIn account and reports tab visibility.
- Admin tools — User management, sender health, usage charts and a monthly invoicing flow.