What is Recruiting OS in one sentence?

Recruiting OS is a productized install of a Hermes-powered AI teammate that lives in your Slack, reads your ATS, and runs five recruiting-specific skills (pipeline overview, candidate research, JD formatting, BD signal scanning, daily recap). It is installed on infrastructure you own, in five business days, for a fixed $4,500.

How is this different from Automindz, Hoxo AI, or other recruiting AI vendors?

Those vendors run the system for you on their infrastructure and charge a monthly platform fee that never ends. Recruiting OS is installed on a VPS you own, with API keys you control, on a Slack workspace under your account. We hand you the keys at the end of week one. Optional retainer is outcome-boxed, not seat-priced, and you can cancel it without losing the system. The economics flip from rent-forever to install-once.

What ATS do you support?

Bullhorn and RecruitCRM ship as Phase 1 connectors. Loxo, Atlas, Greenhouse, Lever, and Workable can be added as a paid extension. Every ATS we support reads and writes through the same universal tool wrapper, which means a new connector is a contained engineering scope rather than a rewrite. If you are on a less common ATS, we will tell you upfront whether the connector adds days to the install.

How is Hermes different from Claude or ChatGPT?

Hermes is an open source agent runtime from Nous Research. It manages persistent memory, tool calls, skill execution, and the Slack and webhook gateways your team interacts with. The model underneath can be Claude Sonnet, Claude Haiku, GPT, or anything OpenAI-compatible. We default to Claude Sonnet for production use. The point of Hermes is the runtime layer, not the model: it is what gives your AI teammate state, durable identity, and the ability to actually do things instead of just talking about them.

Where does my data live?

On the infrastructure you own. The Hermes runtime lives on a VPS you control (we provision one if you do not have one, included in the install). Your ATS data stays in your ATS. Your Slack data stays in Slack. The model API calls send the relevant prompt context to Anthropic or your chosen provider per their commercial terms, which do not train on API traffic. Nothing routes through Moore IQ infrastructure after handoff.

What happens if I cancel the retainer?

The system keeps running. Hermes runs on your VPS. The skills are Markdown and TypeScript files in your repo. The Slack app is under your workspace. Your API keys are yours. The retainer covers ongoing skill updates, prompt drift fixes, and provider API change patches. Without it, you maintain the system yourself or hire someone else to. There is no kill switch on our end.

How is this different from your custom recruiting builds on the /for/recruiters page?

The custom build on /for/recruiters is a six-week scoped engagement priced after an AI Operations X-Ray. Recruiting OS is the productized entry point: a fixed scope, fixed price, five-day install of the five most common recruiting skills. If you outgrow it, you ladder up to a Phase 2 custom build (the Recruiting HQ described below). Most firms start with Recruiting OS and add custom skills over time. The two offers are not in competition. They are the same product road at two different commitment levels.

What is the minimum team size that justifies Recruiting OS?

Five recruiters is the practical floor. Below that, the Slack-channel pattern matters less and the daily recap skill loses value. The economics get sharper at 10-30 recruiters where the daily admin tax compounds across the team. Above 50 recruiters, the standard SKU may need extension. The right rollout is one operator (typically the founder or top biller) running it solo for the first two weeks, then expanding to the desk.

AI Automation for Recruiting Firms

Recruiting OS: a Hermes-powered AI teammate, installed on your stack, owned by you

Productized for 10-50 recruiter firms. Five skills, five days, $4,500. No monthly platform fee. The opposite of renting an Automindz subscription.

Live map

active

System type

overlay

Replacement

not required

Next step

scan

Why renting your recruiting AI stops paying you back

The dominant model for AI in recruiting today is the managed service. A vendor like Automindz, Hoxo AI, or one of the Clay-and-Instantly assemblers builds you a workflow stack on their infrastructure, runs it for you, and bills monthly. The system works. The reply rates are real. The placements happen. And every month forever, you write a check, and the day you stop, the system goes dark.

The math gets worse the longer it runs. A typical managed recruiting AI engagement runs $3,500 to $8,000 per month. At year three, you have paid $126,000 to $288,000 for a capability you do not own and cannot extend. The vendor has the institutional knowledge of your desk: which signals work, which messages convert, which roles your team actually bills. That knowledge lives in their Clay tables and their proprietary scripts. If you leave, you start over.

Recruiting OS flips the model. Same engine, same five skills, same Slack and ATS plumbing, but installed once on infrastructure you own. The system runs whether or not you keep us on retainer. The institutional knowledge accumulates in your repo, not ours. You pay for the install, you optionally pay for ongoing maintenance, and you keep the asset either way. Year three economics for a 20-recruiter firm: roughly $4,500 install plus $0 to $18,000 in optional retainer, versus $126,000 to $288,000 to the managed vendor. The same outcome at ten percent of the lifetime cost, and you own what you paid for.

What is in the box: five skills, one Hermes runtime, your Slack

Recruiting OS is not a chat window. It is a configured Hermes agent runtime running on a VPS you control, connected to your Slack workspace as a bot user and to your ATS as an authenticated API client. The runtime holds memory across sessions, manages tool calls, and runs five recruiting-specific skills out of the box. Skills are Markdown specifications plus TypeScript tool handlers. They live in your repo. You can edit them.

pipeline-overview. Ask the agent in Slack for a pipeline read and it pulls the current state of your ATS: active reqs by stage, candidates sitting in interview, deals stalled past your threshold, replacements due. Default cadence is on demand. A scheduled run can post a snapshot to a Slack channel each morning. The skill works against Bullhorn and RecruitCRM out of the gate, with the connector layer factored so a new ATS adds a single module rather than a rewrite.

candidate-research. Given a job spec or a role title, the agent runs a sourcing cascade against Exa, Apollo, and your selected enrichment provider, scores candidates against the spec, and returns a ranked shortlist with a clear signal stack. This is the same pattern documented on the recruiters pillar: sourcing depth without a LinkedIn Recruiter seat per desk.

JD-formatter.Paste a client’s messy job description into Slack. The agent returns a cleaned posting-ready JD, a derived ICP brief, a Boolean search string for your sourcing tool, and a salary band read for the role and geo. This skill alone tends to save the team between 15 and 25 minutes per new req, and the output is more consistent than a human rewrite.

BD-signal-scan. Run against a target company list (yours, or a list the agent builds from your niche) to identify hiring intent signals: open jobs aging past your relevance window, recent VP-of-talent transitions, funding events in the last 90 days, and the TA-overwhelm signal where the open-req count per existing recruiter crosses a threshold. The skill returns a stacked-signal target list with the evidence per company. This is the Hermes-native version of the Clay table that costs you $500 to $2,000 a month elsewhere.

daily-recap. A scheduled end-of-day post in a Slack channel of your choosing: new candidates entered, replies received, meetings booked, reqs flagged stalled, and the top three priorities for tomorrow. It reads from the ATS and from your inbound channels. The recap exists so the founder or desk lead does not have to assemble it manually at 6 p.m.

How Hermes ties it together

Hermes is the open source agent runtime from Nous Research. It is the layer that gives your AI teammate persistent identity, memory, tool access, and the Slack and webhook gateways your team uses to talk to it. The model underneath is configurable. We default to Claude Sonnet for production work because the quality on structured outputs and tool calling is best in class. You can swap to Haiku, GPT, or anything OpenAI-compatible without rewriting the skills.

The architectural pattern is the same one running in production at one of our existing recruiting clients: a universal tool wrapper in your repo exposes every skill behind a single entry point, the Next.js or FastAPI proxy you already run (or one we install) calls the canonical TypeScript handlers, and the Hermes agent invokes them as needed during a conversation. The benefit of this pattern is zero drift. Your Slack agent, your scheduled cron jobs, and any future client app all hit the same handler code. When a skill improves, every surface improves.

Slack is the default interface because that is where recruiters already live. The agent joins your channels as a bot user. Free-response channels are configured per channel so recruiters do not have to @-mention. Cross-channel routing is configured for the lanes that matter: a dedicated channel for inbound candidate threads, one for BD scans, one for the daily recap, one for ad-hoc questions. The pattern matches the documented agentic coworker pattern: role, memory, and the daily standup test.

Recruiting OS versus the alternatives

The decision most firms actually face is not Recruiting OS versus nothing. It is Recruiting OS versus a managed vendor, versus an enterprise TA platform, or versus rolling your own Clay-and-n8n stack. Here is how the four options compare for a 10-50 recruiter firm.

	Recruiting OS	Managed vendor	Enterprise TA SaaS	DIY Clay + n8n
Year-one cost	$4,500 plus optional $6k retainer	$42k-96k	$60k-500k+	$8k-15k tooling plus your team’s build time
Year-three cost	$4,500 plus optional $18k retainer	$126k-288k	$180k-1.5M+	$24k-45k tooling plus ongoing maintenance
Where it runs	Your VPS, your Slack	Vendor infrastructure	Vendor SaaS	Your stack, your assembly
Cancel and keep using it?	Yes	No	No	Yes
Time to running	5 business days	3-6 weeks	8-16 weeks	Ongoing
Extendable by your team	Yes (Markdown plus TypeScript)	Limited	Vendor roadmap only	Yes, if you have the time

The honest assessment: the managed vendor route works if you do not have technical depth on your team and you want zero operational responsibility. The enterprise SaaS route makes sense at 100-plus recruiters and dedicated TA-ops headcount. DIY makes sense if you have a senior engineer in the room and 200 hours to give them. Recruiting OS sits in the gap most 10-50 recruiter firms actually occupy: you want the productized speed of a managed install without the recurring lock-in or the enterprise overhead.

What the five-day install actually looks like

Day one is provisioning. We stand up your VPS (or work against one you already run), install Hermes and the supporting services, create the Slack app under your workspace and invite it to the channels you nominate, and collect API credentials for your ATS and your chosen enrichment providers. By end of day one, the agent is running on your infrastructure and responding to messages in a sandbox channel.

Days two and three are the skill installs and the integrations. The five skills get configured against your stack: connector wired to your ATS, BD signal sources connected, daily-recap cron scheduled, JD-formatter prompt tuned to your firm’s house style. Each skill is tested end to end against real data from your environment. Anything that does not match your desk gets adjusted before we move on.

Day four is your team. We run a working session with your recruiters: how to use the agent in Slack, what each skill does, what to ask for, what to expect back, where the limits are. This is hands-on. By end of day four, your team has used the agent on at least one real candidate and one real BD scan against their actual desk.

Day five is handoff. You receive credentials for the VPS, the Slack app, the ATS integration, and any API keys we provisioned on your behalf. You receive the repo with the full skill source code and the configuration. You receive a one-page runbook for the most common operational tasks: rotating an API key, adding a Slack channel, editing a skill prompt. If you take the retainer, week two is when the first iteration round happens. If you do not, week two is when your team starts running the system without us.

Phase 2: Recruiting HQ, when the five skills are not enough

Recruiting OS Phase 1 is intentionally bounded. Five skills, Slack interface, ATS read and light write. Most firms run on this configuration for six to twelve months before they want more. When they do, Phase 2 (Recruiting HQ) is the ladder.

Recruiting HQ is a custom Next.js application sitting on top of the same Hermes runtime and the same skill library. Your team gets a dedicated dashboard that visualizes pipeline, flags stalled deals, surfaces the daily recap, and lets your recruiters approve agent actions inline. It runs on the same VPS, against the same data, with the same ownership model. It is the pattern we run for an existing recruiting client today: a production HQ app, deployed via git-push auto-deploy, that turns the agent from a Slack-bot into a full operating dashboard.

Phase 2 pricing starts at $10,000 and scales with the number of custom skills, the depth of the ATS integration, and any custom UI requirements. Three-week delivery, fixed fee, scoped after we see how you actually use Phase 1. If you never want Phase 2, you never need it. Most firms do not. The ladder exists for the ones that outgrow Phase 1, not as a forced upsell.

Security, credentials, and what we never touch

Every credential is yours. The Anthropic (or OpenAI, or OpenRouter) API key is provisioned under your account during the install. The Slack app is created in your workspace under your admin user. The ATS API token is generated against your ATS account. The VPS root credentials are handed to you at handoff. We do not retain copies of any of these unless you specifically request that as part of the retainer for ongoing maintenance access.

The model API traffic follows the provider’s commercial terms. Anthropic’s commercial terms do not train on API traffic. Your candidate data, your client data, your job specs do not get used to improve anyone’s model. The same holds for OpenAI and OpenRouter at the commercial tier. If you have GDPR or CCPA obligations, the install honors whatever consent model your ATS already enforces: do-not-contact flags get respected by every skill, and a re-consent check is added to outreach skills for EU candidates over a configurable age threshold.

Audit logging is built in. Every skill invocation gets logged: which skill, which user, which Slack channel or webhook, which arguments, which result. The log lives on your VPS, in your database. For firms with compliance review, this is the evidence trail.

Pricing, payback, and where the money is

Phase 1 is $4,500 flat for the five-day install. Optional retainer is $500 per month for ongoing maintenance: provider API change patches, prompt drift fixes, uptime monitoring, light skill tuning. The retainer is month-to-month with a 30-day notice. There is no contract minimum. The system runs without it.

Phase 2 (Recruiting HQ) starts at $10,000, scopes per client, optional $750 to $1,500 per month retainer for the larger surface area. Same ownership model. Same cancel-without-loss guarantee.

The payback math is straightforward for a 20-recruiter firm. Daily-recap and pipeline overview save the desk lead roughly five hours per week of manual assembly. JD-formatter saves 20 minutes per new req across the team, which at typical mid-market intake volume is five to ten hours per week. BD-signal-scan replaces a $1,000-2,000 per month Clay table. Candidate-research takes the pressure off a $10,000 per year LinkedIn Recruiter seat. If any one of those four levers lands, the install pays itself back in the first month. If all four land, the install pays itself back in the first week.

What you own at handoff

Everything. The VPS root credentials are yours. The Hermes installation and configuration live on your VPS, not ours. The Slack app is registered in your workspace. The ATS integration uses your API token. The five skills are checked into a git repository you control. The universal tool wrapper, the TypeScript handlers, the Markdown skill specs, the cron schedules, the audit log database schema. All of it is in your repo and runs on your infrastructure.

If you take the retainer, we keep a scoped SSH key and a Slack admin invite so we can do the maintenance work. If you cancel the retainer, you revoke both at the end of the notice period. The system keeps running. This is the same handoff model used in the recruiting sourcing engine case study: you own it the day it ships, and you own it the day we are no longer involved.

Want to see Recruiting OS thinking before you commit? Run the free Resume Anonymizer at /tools/resume-anonymizer. It is powered by the same engine that runs the five Recruiting OS skills. Upload a candidate resume PDF and get back a client-ready anonymized one-pager. Three runs per day, no signup. If the output is the quality you want at scale, the install is the next conversation.

Or run the broader 90-second AI Operations X-Ray for a ranked report on the five highest-leverage automations across your desk, not just recruiting-specific ones.

For recruiters: the full pillar on AI automation in recruiting firms and in-house TA teams
Recruiting sourcing engine case study
AI recruiting outreach personalization
Silver-medal reactivation with recruiting AI
Claude Code for teams
Custom MCP server development

Official resources

Frequently asked questions

What is Recruiting OS in one sentence?: Recruiting OS is a productized install of a Hermes-powered AI teammate that lives in your Slack, reads your ATS, and runs five recruiting-specific skills (pipeline overview, candidate research, JD formatting, BD signal scanning, daily recap). It is installed on infrastructure you own, in five business days, for a fixed $4,500.
How is this different from Automindz, Hoxo AI, or other recruiting AI vendors?: Those vendors run the system for you on their infrastructure and charge a monthly platform fee that never ends. Recruiting OS is installed on a VPS you own, with API keys you control, on a Slack workspace under your account. We hand you the keys at the end of week one. Optional retainer is outcome-boxed, not seat-priced, and you can cancel it without losing the system. The economics flip from rent-forever to install-once.
What ATS do you support?: Bullhorn and RecruitCRM ship as Phase 1 connectors. Loxo, Atlas, Greenhouse, Lever, and Workable can be added as a paid extension. Every ATS we support reads and writes through the same universal tool wrapper, which means a new connector is a contained engineering scope rather than a rewrite. If you are on a less common ATS, we will tell you upfront whether the connector adds days to the install.
How is Hermes different from Claude or ChatGPT?: Hermes is an open source agent runtime from Nous Research. It manages persistent memory, tool calls, skill execution, and the Slack and webhook gateways your team interacts with. The model underneath can be Claude Sonnet, Claude Haiku, GPT, or anything OpenAI-compatible. We default to Claude Sonnet for production use. The point of Hermes is the runtime layer, not the model: it is what gives your AI teammate state, durable identity, and the ability to actually do things instead of just talking about them.
Where does my data live?: On the infrastructure you own. The Hermes runtime lives on a VPS you control (we provision one if you do not have one, included in the install). Your ATS data stays in your ATS. Your Slack data stays in Slack. The model API calls send the relevant prompt context to Anthropic or your chosen provider per their commercial terms, which do not train on API traffic. Nothing routes through Moore IQ infrastructure after handoff.
What happens if I cancel the retainer?: The system keeps running. Hermes runs on your VPS. The skills are Markdown and TypeScript files in your repo. The Slack app is under your workspace. Your API keys are yours. The retainer covers ongoing skill updates, prompt drift fixes, and provider API change patches. Without it, you maintain the system yourself or hire someone else to. There is no kill switch on our end.
How is this different from your custom recruiting builds on the /for/recruiters page?: The custom build on /for/recruiters is a six-week scoped engagement priced after an AI Operations X-Ray. Recruiting OS is the productized entry point: a fixed scope, fixed price, five-day install of the five most common recruiting skills. If you outgrow it, you ladder up to a Phase 2 custom build (the Recruiting HQ described below). Most firms start with Recruiting OS and add custom skills over time. The two offers are not in competition. They are the same product road at two different commitment levels.
What is the minimum team size that justifies Recruiting OS?: Five recruiters is the practical floor. Below that, the Slack-channel pattern matters less and the daily recap skill loses value. The economics get sharper at 10-30 recruiters where the daily admin tax compounds across the team. Above 50 recruiters, the standard SKU may need extension. The right rollout is one operator (typically the founder or top biller) running it solo for the first two weeks, then expanding to the desk.