Juicebox (PeopleGPT)

What it is

Juicebox (branded PeopleGPT) is the AI-native sourcing tool that replaces Boolean search with natural-language queries against an 800M+ profile database compiled from 30+ public sources — LinkedIn, GitHub, conference attendee lists, patent and paper authorship, and other signals. Launched in 2023, it was the first widely-adopted example of LLM-native recruiting tooling, built around a generative-AI primitive rather than retrofitting it onto a legacy product.

As of 2026 it is no longer just a search box. Juicebox now spans the full source → outreach → shortlist loop: PeopleGPT search, multi-step personalized email sequences, and — the headline addition — AI Agents that run search, outreach, and follow-up end-to-end for a continuous pipeline against a standing req. That shift moves Juicebox out of the “better search” bucket and into the autonomous-sourcing conversation alongside SeekOut and goperfect.

Why it shows up in Recruiting stacks

• Natural-language sourcing. “Find me senior Python engineers in Berlin who have contributed to open-source ML libraries and worked at fintech startups” produces a ranked candidate list in seconds. The Boolean queries that used to take a senior sourcer 30 minutes to construct are now a single sentence.
• Multi-source candidate signal. Pulls from LinkedIn (where most candidates live) plus GitHub (depth on technical roles), patent databases, conference speaker lists, and paper authorship — sources a Boolean LinkedIn search can’t reach — and surfaces achievement-level signal, not just titles.
• Autonomous agents for standing reqs. An Agent searches, drafts, sends, and follows up on its own, learning from which profiles get shortlisted. This is the “always-on pipeline” play for evergreen or high-volume technical roles.
• Built for AI-first workflows. Pairs naturally with Claude for outreach personalization and n8n for ATS sync; the assumption is that the recruiter’s stack is itself AI-augmented.

Pricing

Flat per-seat, billed monthly or annually (annual saves ~15%):

• Free — $0, limited trial searches, 1 seat, no contact credits
• Starter — $119/mo, or $99/mo billed annually; 1 seat, unlimited searches, 500 email contact credits/mo, 1 mailbox
• Growth — $199/mo, or $169/mo annually; up to 5 seats, 1,500 credits/seat/mo (email + phone), talent insights, hiring-manager collaboration
• Business — custom; unlimited credits, ATS/CRM integrations, SSO, usage analytics, priority support
• AI Agents add-on — $199/mo per agent, on any paid plan; unlimited contact + email credits, auto-email or auto-shortlist

Real-world cost: the “$99” headline is one seat with no automation. A solo recruiter running Juicebox the way it’s marketed — one seat plus one always-on Agent — is closer to $300–320/mo ($99–119 seat + $199 agent). Budget for the agent line item, not the base price, if the pitch that sold you was autonomous pipeline. ATS/CRM sync sits on the custom Business tier, so integration-dependent teams should price that in early.

Best for

• Growth-stage tech companies hiring engineers, designers, and ML researchers — technical roles where deep, achievement-level signal beats keyword matching
• Recruiting teams already running AI in adjacent workflows (sourcing, screening, interview) who want natural-language search plus an autonomous top-of-funnel
• Teams reducing LinkedIn Recruiter lock-in by moving primary sourcing off the LinkedIn graph

Watch-outs

• Established players out-integrate it on the funnel. hireEZ and Gem — the two incumbents most teams already run — carry deeper ATS/CRM workflow and reporting; Juicebox’s edge is the natural-language UX and the agent layer. Guard: keep your ATS as the system of record and treat Juicebox as the discovery + first-touch layer, not a CRM replacement.
• Sparse-signal roles underperform. Sales, ops, and traditional management candidates leave thin public footprints, so search quality drops. Guard: reserve Juicebox for roles with strong online signal (engineering, research, design) and keep a referral or agency channel for the rest.
• Database depth varies by geography. US/UK/EU coverage is strongest; some APAC and LATAM markets are thinner. Guard: run a 20-candidate test search in your actual target region before committing a seat, and compare recall against Findem or your current tool.
• Autonomous agents send real email on your domain. An auto-email Agent can burn sender reputation or misfire on a bad query. Guard: start any Agent in auto-shortlist mode, review its output for a week, and only switch to auto-email once its shortlists match what you’d have picked by hand.

Alternatives

• hireEZ / Gem — pick these when ATS-native workflow, sequencing, and diversity/pipeline reporting matter more than search UX. They’re the incumbents; Juicebox wins on natural-language discovery, they win on downstream workflow.
• SeekOut — the agentic-slate alternative when you want a vendor-delivered shortlist service (Spot) rather than running the search yourself.
• goperfect — the fastest-moving autonomous-sourcing entrant; weigh it when continuous, multi-channel autopilot outreach across many ATSes is the core need and you’re comparing agent-first products head-to-head.

Not a replacement for an ATS — pair with Ashby, Greenhouse, or another ATS for funnel tracking.