Paraform
Recruiter marketplace for hard-to-fill roles: AI matches a posted role to three to five specialized independent recruiters who source, screen and close it, on a contingency fee of 20–25% of first-year salary paid only on a hire, with a free AI toolkit for the recruiters.
Paraform is a recruiter marketplace for hard-to-fill roles, where a company posts a job and AI matches it to specialized independent recruiters who have filled similar positions before. More than a thousand companies use it, among them Palantir, Rippling, Scale, Thomson Reuters and Abridge, drawing on a network of over ten thousand recruiters across engineering, go-to-market, legal, finance and defense.
The model is outcome-based. Three to five vertical experts work a role in parallel rather than one generalist from a single firm, and they do the recruiting: sourcing, screening, negotiation and closing. A dedicated talent strategist manages the search and acts as the single point of contact for hiring managers. Companies pay a contingency fee of roughly 20 to 25 percent of first-year salary only when a hire is made, in installments across a ninety-day guarantee window, with no retainer and nothing owed if the role stays open. Paraform reports interview-ready candidates in around twelve days on senior searches.
The software sits underneath and on the recruiter side. AI matches roles to recruiters, surfaces good-fit roles to them, prepares intake notes from hiring-manager calls so they start aligned, and handles the follow-ups, client communications and payments. Recruiters get an all-in-one toolkit with sourcing, candidate matching, a CRM, a notetaker and a scheduler, free of charge, and earn a 70 percent split of the fee. Companies get real-time dashboards over the pipeline, with collision prevention and submission limits so several recruiters working one role does not become noise.
Paraform also says its agents learn from every search, interview and hire, so what a particular company considers a strong candidate informs later matching. It is SOC 2 certified, with a published trust portal. Founded in 2022 and based in San Francisco, it raised a Series B in March 2026 led by Scale Venture Partners.
Vendor details
Canonical URL
https://www.paraform.com
Category
Enterprise operations agent
Subcategory
Agentic hiring recruiter marketplace
Funding status
Independent, based in San Francisco, founded in 2022 by co founders John Kim and Jeffrey Li, who first recruited each other and concluded the recruiting industry was broken. Paraform grew to eight figures in annual recurring revenue with a roughly twenty person team and raised a Series B in March 2026 led by Scale Venture Partners, with participation from Felicis, A star, and Liquid 2 Ventures, plus founders and executives from Palantir, Canva, YouTube, Stripe, Shopify, and Uber, bringing total funding to sixty five million dollars. More than one thousand companies, including Palantir, Rippling, Decagon, Abridge, and Cognition, hire through the platform.
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
Paraform's software is largely first-party rather than connective: recruiters on the network receive a free all-in-one AI toolkit bundling sourcing, candidate matching, a CRM, a notetaker and a scheduler, and Paraform runs client communications, feedback and payments through the platform itself. The notetaker captures hiring-manager calls and the scheduler books interviews, so meeting and calendar surfaces are reached in practice, but no external system is named anywhere on the pages read and there is no integrations page, connector directory or count. Integration with a company's applicant tracking system, email and calendar is not confirmed on the vendor's pages. No developer API, SDK or MCP surface is published.
In practice
Your hardest engineering and go to market roles sit open for months while internal sourcing stalls. Paraform matches you with specialized recruiters and custom AI agents that deliver interview ready candidates in roughly ten to twelve days.
Traditional agencies charge retainers and flood you with mismatched resumes. Paraform pairs AI matching with vetted recruiters, keeps quality gates and submission limits in place, and you pay a bounty only when you hire.
You want AI to speed up hiring without losing the human judgment that actually closes candidates. Paraform's agents handle sourcing and busywork while recruiters and your team stay in the loop on every decision.
Sources & related URLs
Agentic Index coverage score
4.0 / 14 capabilities · 29%
| Integrations & Tool Calling | Partial |
|---|---|
|
The homepage names no external system: it names customers (Palantir, Rippling, Decagon, Abridge, Scale, Thomson Reuters) and no software, and the site has no integrations page, connector strip, directory or count. Integration with a company's applicant tracking system, email and calendar is not described. The tooling is Paraform's own. The recruiter toolkit bundles "sourcing, candidate matching, CRM, notetaker, scheduler and more, completely free of charge". A notetaker joins meetings and a scheduler books them, so calendar and meeting surfaces are reached in practice even though no provider is named, and payments flow through the platform, since Paraform manages client comms through to feedback and payments. Those are real actions into surfaces Paraform does not own. The tools are mostly first party substitutes rather than connectors: Paraform ships its own CRM rather than integrating with one, so the integration estate is narrow by design. Sourcewww.paraform.com homepageread 2026-09-12 |
|
| Workflow Orchestration | Partial |
|
The agents assist a human workflow rather than running one. What they do is real: a role is posted and AI "matches it to recruiters who have deep expertise in that specific hiring vertical", three to five working it in parallel; the agents then do the heavy lifting by surfacing great fit roles, getting hiring managers interested and handling the busy work, and the toolkit's notetaker prepares intake notes from hiring manager calls. Those are several named steps with distinct participants: the company, the matching agent, the recruiters, the hiring manager and the talent strategist. The consequential sequence is carried out by people. Recruiters "source, screen, and deliver candidates" and handle negotiation and closing, and a dedicated talent strategist actively manages the search. No orchestration runtime is named, no handoff between agents is described, no plan is built per request, and there is no authoring surface of any kind. Paraform's own framing says as much: "Most services in this space give you either a tool or a person. Paraform gives you both." Custom AI agents run sourcing and matching workflows alongside specialized recruiters, but the journey from role to placement is one the recruiters walk. Sourcewww.paraform.com homepage, /insights/paraform-vs-recruiting-agencies and /blog/best-platforms-hiring-remote-engineering-teamsread 2026-09-12 |
|
| Knowledge Grounding & RAG | Partial |
|
The one retained artifact is the intake note. Recruiters "receive pre-prepared intake notes from hiring manager calls, so they are aligned from day one", captured by the toolkit's notetaker and kept between searches for the same company. Candidate matching and sourcing run in the recruiter toolkit over the marketplace's own candidate pool. The learning claim is model learning, not a corpus: "We learn what great looks like for you. That knowledge compounds." A model that has learned is not a store that is retrieved. No index, corpus size, search surface, query language or retrieval mechanism is published, and no page states that anything persists between runs or is queried per request. Paraform publishes a recruiter network size of more than 10,000 and no candidate corpus at all, because its candidates arrive through recruiters rather than out of an index. No integration with a company's applicant tracking system, which would bring the company's own pipeline into scope, is described. Sourcewww.paraform.com homepage and /blog/paraform-vs-bountyjobsread 2026-09-12 |
|
| Human Oversight & Guardrails | Partial |
|
The people in the loop are the product, so they are not a control over the agents. Paraform is a curated network of more than 10,000 independent recruiters matched three to five per role, and those recruiters "source, screen, and deliver candidates who are actually interview-ready". Human recruiters handle negotiation, cultural fit and passive candidate outreach, while AI handles matching, workflow management and matching intelligence behind the scenes. On top sits white glove support, with a dedicated talent strategist actively managing the search. A recruiter reviewing a candidate is a person doing the job, not a person approving an agent. What the platform does provide is real. Decisions stay with the company: the agents surface roles, match recruiters to roles and prepare intake notes, and the company decides whom to interview and hire. The marketplace also ships bounded controls on what reaches a buyer, in quality gates, collision prevention and submission limits, which cap the volume and duplication a company sees. Nothing documents a configurable hold over an agent's own action. The agents' consequential outputs are surfacing and matching rather than contacting candidates, so there is no approval step before an agent acts, because the agent is not the one acting. Sourcewww.paraform.com homepage, /insights/paraform-vs-jack-and-jill and /insights/paraform-vs-recruiting-agenciesread 2026-09-12 |
|
| Security, Identity & Governance | Partial |
|
Paraform handles candidate and applicant tracking data and holds SOC 2, but publishes no customer facing access controls. The footer of every page reads SOC 2 Certified, linked to a live trust portal hosted by Vanta at app.vanta.com/paraform/trust. Because the compliance vendor maintains that portal rather than Paraform, it is stronger evidence than a static badge. No SSO, RBAC, SCIM, permission model, admin console or access governance control appears on the homepage, and there is no security page in the navigation: the SOC 2 link in the footer is the whole security surface the site presents. In the marketplace model, candidate and hiring data is visible to independent third party recruiters who are not the buyer's employees, three to five per role, and nothing on the site describes how that access is scoped, logged or revoked. The SOC 2 scope is not stated on the site itself. Sourcewww.paraform.com footer and homepageread 2026-09-12 |
|
| Observability & Auditability | Partial |
|
Paraform shows the hiring pipeline, not what its agents did. Real time dashboards, quality gates, collision prevention and submission limits keep companies in control and give visibility into the pipeline, but each is a marketplace control over what human recruiters submit: collision prevention stops two recruiters claiming the same candidate, submission limits cap how many each may send, and dashboards show the state of the hiring funnel. The dashboards do give a company a live operational view of submissions and stage. The agents do produce one inspectable artifact. Intake notes are prepared from hiring manager calls by the toolkit's notetaker and read later by recruiters, so a reader can see what the agent captured and from which conversation. Nothing keeps a record of agent conduct that could be reconstructed. No page names a run log, action history, audit trail, export or retention period, and nothing states why the matching agent paired a given recruiter with a given role, which candidates it surfaced and passed over, or on what signal. That matters here because matching recruiters to roles is the agent's central act, and it decides who works a company's search. Sourcewww.paraform.com homepage and /blog/paraform-vs-bountyjobsread 2026-09-12 |
|
| Memory & State Persistence | Not documented |
|
What Paraform describes is model learning, not agent memory. In the vendor's words: "With every search, interview, and hire, we learn what great looks like for you. That knowledge compounds and powers AI agents that help you throughout the hiring process." Learning that compounds into a matching model improves the system, but it is not a store a customer can point at. Paraform's comparison material calls it "a compounding intelligence layer that gets sharper with scale", built from "AI that learns from every search, interview, and placement across the system": pooled across customers, which makes it a model rather than memory kept for each customer. Nothing on the site describes a named memory layer, a retained preference set a customer can inspect, decision traces carried between searches, a stated scope or lifetime, or a control to inspect, edit, export or purge agent state. Intake notes prepared from hiring manager calls do persist and are read by recruiters on later searches; they work as knowledge rather than memory. Sourcewww.paraform.com homepage and /insights/paraform-vs-doverread 2026-09-12 |
|
| Deployment & Data Residency | Not documented |
|
No hosting, region or residency statement is published. Paraform is delivered as a cloud marketplace, and nothing about region, residency, data location, hosting or environment appears on the homepage, in the navigation or in the footer; the vendor does not make even a one sentence data location claim. There is no security page and no trust or infrastructure content on the site itself: the whole security surface is a SOC 2 Certified link in the footer. The only geographies published are market coverage. Paraform operates across the US, Canada, UK and Australia, with recruiters in defense, legal, healthcare and quant verticals. That is where it recruits, not where data sits, and Paraform Inc. being a San Francisco company says nothing about infrastructure. This matters more than usual for a marketplace. Candidate data is handled by independent third party recruiters in several countries, not only by the vendor, so where it is processed is a live question the site does not address. Any hosting region, subprocessor list or residency commitment would sit in the Vanta trust portal at app.vanta.com/paraform/trust. Sourcewww.paraform.com homepage and footerread 2026-09-12 |
|
| Prebuilt Agents, Templates & Packs | Partial |
|
Recruiters get a bundled toolkit with five named working components, shipped ready and free: "an all-in-one AI toolkit with sourcing, candidate matching, CRM, notetaker, scheduler and more, completely free of charge." Each does distinct work, a recruiter joining the marketplace has them working on day one without building anything, and the notetaker and scheduler in particular do work when invoked. The agents that serve companies are custom built, not prebuilt. Paraform's framing is "custom AI agents that think like you", assembled per company from what it has learned from that company's searches, so a buyer does not choose among ready agents; they receive ones made for them. The toolkit components are modules of one product rather than agents a buyer picks among: sourcing feeds matching, which feeds the CRM, and nothing can be adopted or removed separately. No catalog exists in any form: no library, marketplace, template gallery or install path appears anywhere on the site. Sourcewww.paraform.com homepageread 2026-09-12 |
|
| Triggers & Channel Coverage | Partial |
|
The agents act without being asked, but no outbound channel is named. They "surface great-fit roles" to recruiters, "get hiring managers interested" and "handle the busy work", and Paraform says it manages client comms, feedback and payments and tells recruiters to "leave the follow-ups to us". Follow-ups the platform owns are messages the agent initiates. Events drive the flow at two named points: a posted role triggers AI matching to recruiters with relevant vertical expertise, and a hiring manager call triggers the notetaker, which produces the intake notes recruiters read. The scheduler books interviews, which is a calendar event the platform creates. No email, SMS, WhatsApp, phone or in app messaging channel is named. The only named surface is Slack, and it belongs to the dedicated talent strategist, a person, rather than to the platform. Triggering is not broadly real time or multichannel, and no schedule control, cadence setting or event framework reaching outside the product is documented. Sourcewww.paraform.com homepageread 2026-09-12 |
|
| Model Flexibility & Routing | Not documented |
|
Nothing tells a customer which model runs, and nothing lets them choose one. The homepage speaks of "custom AI agents that think like you" and an AI toolkit with sourcing and candidate matching, and Paraform's comparison pages describe AI that "matches roles to recruiters" and "learns from every search, interview, and placement across the system". No model or provider is identified for any of it. There is no model picker, admin setting, per agent assignment or bring your own key path on the homepage, the navigation or the footer, and no settings or configuration surface is published at all, which follows from the product being sold as a managed marketplace rather than a self serve platform. The custom in custom AI agents is not model choice. The agents are tuned to a company's hiring preferences, learned from its searches, which shapes the agent's behavior rather than letting the customer choose what runs underneath. Sourcewww.paraform.com homepage and /insights/paraform-vs-doverread 2026-09-12 |
|
| APIs, SDKs & MCP Extensibility | Not documented |
|
No outside system or agent can call Paraform through anything published. Neither the site navigation nor the footer has a developer, docs, API or integrations entry, no docs or dev subdomain is published, and there is no llms.txt or page for assistants. No API reference, endpoint list, authentication guide, SDK, webhook or MCP server appears anywhere. The For connectors section is not a developer surface despite its name: it serves a third participant type in the marketplace, people who refer candidates or clients. The commercial shape fits the absence. Paraform sells outcomes on contingency, with recruiters and a talent strategist as the interface and a free toolkit as the software, so it has little reason to publish a build surface, and there is no self serve tier a developer could reach. An enterprise API could still sit behind a sales conversation without being published. Sourcewww.paraform.com homepage, navigation and footerread 2026-09-12 |
|
| Testing, Debugging & Optimization | Not documented |
|
Nothing measures the agents themselves. The published metrics measure the service: a 70 percent interview rate, 3x faster to hire, 98 percent satisfaction and roughly twelve days to hire are outcomes of a marketplace in which human recruiters do the sourcing, screening, negotiation and closing, so they say nothing on their own about the matching agent. The placement guarantee is a commercial warranty. A ninety day window on a human placement moves risk to the vendor, which is a good term for a buyer, but it does not measure agent behavior. Quality gates, collision prevention and submission limits are marketplace controls over recruiter submissions. No test environment, sandbox, holdout, versioning, regression check, evaluator or scored result is published, and there is no live metric on the agents' own output: no match acceptance rate, no surfacing precision, nothing showing whether the matching agent is getting better or worse. No fairness audit of model outputs across demographic groups is described. The one learning claim, that knowledge "compounds and powers AI agents", belongs to Paraform Intelligence and comes with no measurement attached. Sourcewww.paraform.com homepageread 2026-09-12 |
|
| Browser & Computer Use | Not documented |
|
No browser or computer use; the product's shape rules it out. The agents' work is matching, surfacing and note taking: AI matches roles to recruiters with relevant vertical expertise, agents surface great fit roles and get hiring managers interested, and the toolkit's notetaker captures hiring manager calls. None of that involves operating an interface. No hosted or local browser, desktop session, remote or local computer control, RPA or extension appears on the homepage, the navigation or the footer. The sourcing is human. It is done by more than 10,000 independent recruiters using their own networks and tools, with Paraform's toolkit assisting them, and there is no sourcing browser extension. A recruiter browsing LinkedIn on their own account is a person doing their job with no Paraform agent involved, and the platform makes no claim over it. Nothing Paraform does depends on what is drawn on a screen. Sourcewww.paraform.com homepage and /insights/paraform-vs-jack-and-jillread 2026-09-12 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Recent platform changes
Paraform introduced Hiring Intelligence, a foundational data layer designed to power every AI agent across its platform. This new capability leverages years of real-world hiring data to enhance agentic workflows and candidate matching.
Bears on: Agent capability
View sourcePricing
No software price; 20–25% of first-year salary on placement, nothing if no hire is made
A contingency placement commission, charged as a percentage of the hired candidate's first-year salary and payable only on a hire, with installments across a ninety-day guarantee window. There is no software subscription on either side: the recruiter toolkit is free of charge, and recruiters are paid a split of the commission.
What is public
The commercial model is published in prose, though not on a pricing page, verified first-party 2026-09-12. Paraform charges companies a contingency placement fee of roughly 20 to 25 percent of the hired candidate's first-year salary, with around 25 percent cited as typical, payable only when a hire is made and with zero cost if no hire results. There are no retainers and no hourly billing. Payment is by installments across a ninety-day guarantee window. On the recruiter side of the marketplace the all-in-one AI toolkit, covering sourcing, candidate matching, CRM, notetaker and scheduler, is completely free of charge, and recruiters earn a 70 percent split of the fee against a stated 50 percent at traditional agencies; Paraform states it has paid out over $50M to recruiters on the network. Scale figures are published alongside: 10,000-plus expert recruiters, 1,000-plus companies, a 70 percent interview rate, three times faster to hire and roughly twelve days to hire on senior roles. The site offers both a self-serve signup and a demo route, with no published rate card behind either.
Billing mechanics
Outcome based contingency model: a placement fee, roughly twenty to twenty five percent of first year salary, is charged only when a hire is made, with no retainers and a ninety day guarantee. The recruiter side toolkit is provided free.
Cost watchouts
Cost scales directly with hires and salaries, so a high volume of senior placements adds up, and the contingency percentage is higher than some lower touch software only sourcing tools.
Variable cost rationale
Raised from medium, and the reasoning is inverted from the July version rather than adjusted. On a contingency model there is no floor and no ceiling: nothing is owed while no hire is made, and each hire triggers a fee of roughly a quarter of that person's first-year salary. A single senior placement at a $250,000 salary costs more than a year of any subscription product in this sublane, and a team hiring ten roles carries six-figure exposure that no plan caps. The customer controls the trigger, which is the mitigation — cost tracks hires rather than activity, so a paused hiring plan costs nothing and there is no contract to exit. But exposure per event is an order of magnitude above anything metered per seat or per candidate contacted, and the 90-day guarantee window transfers some risk back without capping the fee. Graded high on magnitude with the customer's control over the trigger named.
Additional watchouts
This is a commission, not a license, and the comparison that matters is against an agency rather than against software. At 20 to 25 percent of first-year salary the fee lands just inside the SHRM band Paraform itself cites for headhunters, so the saving is in speed, parallelism and the absence of a retainer rather than in the headline percentage — the pitch is three to five specialized recruiters working a role at once with nothing owed if nobody is hired. Confirm three things in writing: where in the band your roles sit and what moves them, what the ninety-day guarantee actually pays or replaces if a hire leaves, and how the installment schedule interacts with that window. Two structural points a software buyer should not miss: there is no subscription to cancel and therefore no ongoing cost when hiring pauses, and the candidate and hiring data passes through independent third-party recruiters who are not your employees, three to five per role, with nothing published about how that access is scoped or revoked. Finally, note that the free AI toolkit is for recruiters, not for you.
Sales call required
Yes, required for paid access
Free / trial
No upfront cost to post roles; companies pay only on a successful hire, and the recruiter toolkit is free
Lowest paid plan
Contingency placement fee around twenty to twenty five percent of first year salary per hire
Commercial notes
Positioned as a recruiter marketplace with AI agents rather than a subscription software tool, with the contingency percentage below traditional full service agencies but above pure software only sourcing tools.
Key ambiguities
The pricing is published, but it is not software pricing. Paraform charges a placement commission of roughly 20 to 25 percent of first year salary, payable only on a hire, in installments across a ninety day guarantee window. There is no subscription, seat price or usage meter on either side of the marketplace, so there is no entry price to list; the actual terms are set out in the profile. The free tier belongs to the other side of the market. The all in one AI toolkit is free to recruiters, who then earn a 70 percent split of the commission against a stated 50 percent at agencies. A company buyer gets no free tier; it gets recruiters. Paraform's own comparison pages state the ninety day guarantee window plainly. Against a subscription tool, the cost of a contingency model is unbounded per hire but zero when hiring stops, which inverts the usual risk profile of software.
Cancellation / refund
No retainers; companies pay a contingency fee on a successful hire, often with a ninety day guarantee that covers a replacement if a hire does not work out.
Missing data
No pricing page exists in the navigation or the footer, so the terms are assembled from Paraform's own comparison and insights pages rather than from a rate card. Unpublished: where in the 20 to 25 percent band a given role lands and what moves it, whether the percentage is negotiable at volume, the installment schedule inside the ninety-day window, what the guarantee pays out if a hire leaves inside it, whether enterprise agreements differ, and whether the Paraform Intelligence product announced on the blog is separately priced. On the recruiter side, the split is stated at 70 percent but the terms around back-office support, payment timing and any deductions are not.
Related vendors
- 11th Estate — Agentic platform whose AI engine scans markets, matches an…
- Adopt AI — AI CPA firm whose agents run reconciliations, close, AP and AR, and…
- Aera Technology — Decision intelligence platform where an always on agent executes…
- Akro AI — On premise operational intelligence that automates document heavy…
- alfred_ — AI executive assistant that triages the inbox overnight and scores…
- Alloy.ai — Commerce intelligence system for consumer brands that unifies four…
Alternatives to Paraform
The closest documented capability profiles to Paraform among enterprise operations agents tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.
- effie.ai4.5 / 14Fuller documented coverage on Knowledge Grounding & RAG
- Casca5.0 / 14Fuller documented coverage on Security, Identity & Governance and Triggers & Channel Coverage
- HackerEarth4.0 / 14Adds documented APIs, SDKs & MCP Extensibility
- Norm Ai5.0 / 14Fuller documented coverage on Knowledge Grounding & RAG and Human Oversight & Guardrails
- Synthetic3.0 / 14A lighter documented profile than Paraform
- Balerion AI4.5 / 14Fuller documented coverage on Workflow Orchestration and Knowledge Grounding & RAG
Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded