TLDR
- On Sept. 14, 2026, Workable put four AI recruiting agents into general availability for paid accounts worldwide: Job Brief, Sourcing, Screening, and Engagement.
- The agents are built into Workable’s ATS and priced on a credit model (3,000 free credits on every plan; screening 1 credit, sourcing 2, completed candidate chat 10).
- A new Workable MCP Server connects recruiting and HR actions to Claude, ChatGPT, and other Model Context Protocol clients at no extra cost, with existing Workable permissions enforced.
- Workable-reported early results: agents have processed 120,000+ candidates; Belong’s 3-person team filled a Head of Operations role in seven days; CacheFly’s Martin Lagler said Screening “eliminated about 20% of unqualified applications automatically.”
- For buyers: this is ATS-native agentic recruiting (source → screen → engage), not the same category as Amazon Connect Talent-style AI voice interviews. Ask about credit burn, Ideal Candidate Profile governance, and audit trails before treating GA as a blank check.
Workable announced global general availability of its AI Recruiting Agents on Sept. 14, 2026. The company frames the launch as a shift from recruiter-driven busywork to agent-run pipeline work, with humans keeping ownership of relationships, calibration, and hiring decisions.
That is useful for HrTechSaas readers because mid-market and growth TA teams already live in ATS workflows. The news is not another standalone sourcing chatbot. It is four agents wired into the same system of record, plus a credit meter and an MCP bridge into the AI assistants many recruiters already use.
This piece stays on Workable’s GA claims, pricing, and buyer questions. It is not a rewrite of high-volume AI recruiting assistant listicles or AI interview tool roundups already on the site.
What shipped in GA
Per Workable’s Sept. 14 product post, four agents work as one workflow:
- Job Brief Agent structures hiring requirements from the role.
- Sourcing Agent continuously reaches passive candidates.
- Screening Agent scores every applicant against a structured Ideal Candidate Profile across 14 dimensions.
- Engagement Agent opens a personalized chat when information is missing, then re-scores after answers land.
Workable says recruiters can attach agents to jobs already in progress, adjust the Ideal Candidate Profile while a job is live (changes apply to recently processed candidates), and see each candidate’s agent status and planned next steps. Interruption warnings appear before an action would disrupt an active agent workflow.
The practical output Workable promises is a ranked shortlist of qualified, interested candidates ready for human interviews, with retained context across screening drops, re-scores, and credit use by job.
Credit pricing buyers need in writing
Workable’s GA pricing is explicit in the same post:
| Item | Credit cost / price |
|---|---|
| Included on every Workable plan | 3,000 free AI credits |
| Screen one candidate | 1 credit |
| Source action | 2 credits |
| Completed candidate chat | 10 credits |
| 5,000 credit bundle | $0.12 per credit |
| 10,000 credit bundle | $0.10 per credit |
| 50,000 credit bundle | $0.095 per credit |
Agents are a paid add-on to Workable Recruiting plans. An AI Credits Report breaks consumption by job, action, and time period, with filters for department, location, and job status.
For finance and TA ops, that meter matters more than the feature list. At 10 credits per completed chat, engagement volume can dominate spend even when screening looks cheap. Model burn against your typical application volume and outreach cadence before you enable agents on every open req.
MCP Server: agents outside the ATS UI
Workable also launched a Workable MCP Server with GA. It connects recruiting and HR capabilities to Claude, ChatGPT, and other MCP-compatible assistants so teams can source candidates, update pipelines, manage employee records, and run reports from chat without developer setup.
Workable states the MCP Server:
- Is included at no extra cost on every subscription plan
- Respects existing Workable user permissions
That is a different product story from “AI inside the ATS only.” It puts Workable into the same agent-interoperability conversation as other HCM vendors shipping MCP or connector strategies. It also raises a governance question: who can instruct Claude to update a pipeline, and how is that action audited back to a named user?
Early results Workable published
Workable attributes these outcomes to early agent use (all vendor-reported):
- Agents have processed more than 120,000 candidates since initial rollout
- Early adopters named: CacheFly, Belong (UK dementia care), H2 Health, HSI
- Belong’s 3-person recruiting team filled a Head of Operations role in seven days
CacheFly head of HR Martin Lagler said the Screening Agent “eliminated about 20% of unqualified applications automatically. In the past, I had to go through every answer manually. Now I don’t waste anyone’s time.”
H2 Health director of recruiting Carol Ribeiro said of continuous sourcing: “Anything that’s doing the sourcing in the background saves us time. It’s less time that the recruiter actually needs to be sourcing for that specific position.”
Treat quotes and case metrics as directional. Ask for baselines (time-to-shortlist, screen-reject rate, credit spend per hire) in roles that match your volume and compliance profile.
How this differs from existing HrTechSaas coverage
The site already covers adjacent buying jobs:
- Best AI sourcing tools and LinkedIn Recruiter alternatives
- Best AI recruiting assistants for high-volume hiring
- Paradox vs Phenom vs Humanly
- Best AI interview tools and HireVue alternatives
Those pages help buyers shortlist categories. This draft is different:
- News explainer on Workable Agents GA (Sept. 14, 2026)
- Focus on four-agent workflow, credit economics, and MCP
- Not a multi-vendor sourcing listicle and not an AI interview product comparison
Do not merge this into the high-volume assistant or HireVue-alternatives URLs. Different intent, different publish job.
It is also distinct from Amazon Connect Talent-style AI voice interviewing. Workable’s GA story is ATS-native source/screen/engage agents. Interview intelligence and automated interview scoring remain a separate evaluation, covered in the HireVue alternatives guide and related AI-in-HR explainers.
Who should care, and what to ask
Prioritize a read-out if you:
- Already run Workable Recruiting and want agentic sourcing/screening without adding a second ATS
- Are comparing mid-market AI recruiting assistants and want a credit-metered alternative to seat-only pricing
- Need MCP or chat-side pipeline actions under existing permission models
Ask Workable (and your legal/compliance partner):
- Which agent actions write back to the candidate record versus recommend only?
- How is the Ideal Candidate Profile versioned, and who can change scoring weights mid-campaign?
- What audit trail exists for Screening scores, Engagement questions, and MCP-initiated updates?
- How do credit burn and the AI Credits Report map to cost-per-qualified-candidate in your volume bands?
- What candidate notice, consent, and adverse-impact testing documentation ships for jurisdictions that regulate automated employment decision tools?
- Can candidates opt out of Engagement chat without automatic reject, and how is that logged?
The take: GA is real, unit economics decide adoption
Workable’s Sept. 14 GA is concrete product news: four agents, global paid-account availability, published credit prices, and an MCP server included on every plan. For operators, the durable story is whether credit burn and Ideal Candidate Profile quality beat the cost of manual screen-and-source work without creating compliance debt.
If you already live in Workable, this is the moment to pilot on a bounded set of reqs, watch the AI Credits Report weekly, and lock scoring ownership before you turn agents on for every open role. If you are evaluating across Paradox, Phenom, Humanly, and ATS-native agents, put Workable on the shortlist for teams that want sourcing and screening inside one ATS rather than another point tool. Feature breadth is not the test. Traceable scores, permissioned MCP actions, and predictable credits per hire are.
FAQ
When did Workable Agents go generally available?
Workable published the GA announcement on Sept. 14, 2026, stating the four AI recruiting agents are available to paid Workable accounts globally.
Which agents are included?
Job Brief Agent, Sourcing Agent, Screening Agent, and Engagement Agent.
How does credit pricing work?
Every plan includes 3,000 free AI credits. Screening costs 1 credit per candidate, sourcing costs 2 credits, and a completed candidate chat costs 10 credits. Additional credits sell in 5,000 / 10,000 / 50,000 bundles at $0.12 / $0.10 / $0.095 per credit.
Are the agents free?
No. Agents are a paid add-on. The MCP Server is included at no extra cost on all subscription plans.
What is the Workable MCP Server?
A Model Context Protocol server that lets Claude, ChatGPT, and other MCP clients run recruiting and HR actions in Workable while respecting existing user permissions.
How many candidates have the agents processed?
Workable reports more than 120,000 candidates processed as of the GA post.
Is this the same as AI interview tools like HireVue alternatives?
No. Workable’s GA focus is ATS-native job brief, sourcing, screening, and engagement agents. AI-led interview products are a separate category.
