ProductAI agents

Agents that actually run your recruiting.

Seven first-class agents for sourcing, screening, scheduling, outreach, references, triage, and reporting. Bring your own Claude or ChatGPT, or use ours. Every action is auditable. Humans approve before anything ships.

Start for freeBook a demo
01The agents/ seven, first-class

Each one does one thing extremely well.

Named, scoped, and shaped to a specific recruiting job. You decide which to switch on, in what order, and where the human stays in the loop.

S
Sourcing agent
Always on

Natural-language search across 400M+ profiles in 30+ data sources, ranked against your role rubric.

OutputScored shortlist, deduped, with reasoning per candidate.
GateRecruiter approves before any outreach.
R
Screening agent
Voice + chat

Runs first-round calls or chat screens. Scores communication, role fit, and signal across a structured rubric.

OutputRecording, transcript, summary, score, and recommended next step.
GateHiring manager reviews before advance.
O
Outreach agent
Multi-channel

Picks the best channel per candidate, personalizes off public profile data, and tracks every reply.

OutputDrafted messages across email, LinkedIn, and SMS.
GateDrafts queued for recruiter approval by default.
C
Scheduling agent
Calendar + panel

Coordinates candidate, panel, and timezone. Books, reminds, reschedules, and sends prep materials.

OutputCalendar invites, agendas, and pre-interview briefs.
GateAuto-confirms within recruiter-defined limits.
F
Reference agent
Synthesis

Emails referees with consent, structured questions, and summarizes the responses.

OutputSynthesized feedback with strengths, concerns, and consensus signal.
GateConcerns flagged for human review.
T
Triage agent
Inbound

Scores every inbound application against the JD, deduplicates, and routes to recruiter or talent pool.

OutputStage advancement, polite decline, or talent pool nurture.
GateDecline messages always recruiter-approved.
D
Reporting agent
Daily digest

Summarizes the week's pipeline movement, flags stalled candidates, and recommends concrete next actions.

OutputSlack digest, dashboard update, suggested nudges.
GateRecommendations require approval before send.
02How they run/ live

One chat. The whole motion.

A real agency-recruiter sequence playing out the way it would in your workspace. Type what you want. The agents source, draft, send, track, and book. You stay in control of every send.

Aira
AI Recruiting Assistant
Online
⌕Search candidates, create a job…↑
Reasoning chain

Every step is shown. Every decision is cited.

Watch the sourcing agent reason through a real query. Search, filter, rank, and explain. No black box.

NNew candidate found
Sarah Johnson · Senior Backend Engineer · Berlin
⌘Verify experienceAI
Does the candidate have 5+ years of Rust in production?
Yes, lead engineer on Rust services at 2 prior companies
⌘Assess fitAI
Is the candidate likely open to a move?
High intent · 3y in role, posted job preferences last week
⌘Compose outreachAI
Personalized message referencing their open-source work
Sent · awaiting reply
03Bring your own AI/ no lock-in

Whatever model you use today, and whatever you switch to.

Vitae is the first ATS built for the agent era. Native MCP server, complete read and write API, model-agnostic execution. Your AI choice should never be a function of your ATS.

MCP

Native MCP

Every Vitae object, field, workflow, and agent action is exposed as an MCP tool. Your Claude or ChatGPT drives the platform without any glue code.

BYO

Bring your keys

From Solo upwards, plug in your Anthropic, OpenAI, Cursor, or self-hosted keys. We never mark up token costs. You pay providers directly.

Routing

Switch per agent

Use Sonnet for screening, GPT-4o for outreach, a fine-tune for scoring. Configure per agent, per workflow, or globally.

curl·POST /v1/mcp/agents/source200 · 38ms
{
  "agent": "sourcing",
  "model": "claude-sonnet-4",
  "query": "senior backend, Berlin, Rust 5y+",
  "approval_gate": "human",
  "max_results": 25
}
04Trust/ approvals + audit

Agents that do not act unilaterally.

Every agent runs through a configurable approval gate. Every action is logged, attributed, and reversible. Auditable by default.

92K
Candidates screened by Vitae agents
100%
Outreach goes through human approval by default
4.9/5
Recruiter satisfaction across active workspaces
0
Tokens marked up. You bring your own keys, you pay providers direct.
“Vitae cut our sourcing time by more than half. It feels like having an extra recruiter on the team. The shortlists are consistently on point.”
NF
Nima Farighi
Managing Director, Grosvenor Search
05Agent FAQ/ six questions

Questions about the agents.

The questions we get most often from teams trying out the AI agents.

By default, our managed agents run on Claude Sonnet and GPT-4o, picked per task for speed and quality. From Solo upwards, you can connect your own Anthropic, OpenAI, Cursor, or self-hosted model keys and we will route every agent call through them.
Related reading

The case for agentic recruiting.

Why agents — not features — are the unit of progress in modern recruitment software.

Agentic AI Rise: Transforming Recruitment After CES 2026
AI News

Agentic AI Rise: Transforming Recruitment After CES 2026

March 11, 2026 · 6 min read
AI Recruitment Software Should Run Hiring, Not Just Track It
Recruitment AI

AI Recruitment Software Should Run Hiring, Not Just Track It

April 29, 2026 · 6 min read
Top AI-Powered Recruitment Tools: Platforms Every HR Team Should Know
Recruitment AI

Top AI-Powered Recruitment Tools: Platforms Every HR Team Should Know

April 8, 2026 · 8 min read

Put the agents to work.

Switch them on one at a time. Approve every send. Watch your time-to-fill drop.

Start for freeBook a demo