Recruiting software that Claude, ChatGPT, and Cursor can drive.
Kit is an applicant tracking system (ATS) with 160+ MCP tools, 6 specialized agents, and your choice of Gemini, Anthropic, or OpenRouter keys.
Kit publishes more than 160 MCP tools, each labelled read-only, destructive, or reaches a real person before your client runs it. Candidates need no account and no password, only a magic link. Or run Kit's own six agents instead.
Every ATS says 'AI-powered.' Here's what they mean.
Template text. Black-box models. Zero integration with your actual AI tools.
Auto-generated job descriptions. The same ones everyone else gets.
Most ATS vendors call a single GPT endpoint to rewrite your job post. That's their entire AI feature. You could do it yourself in ChatGPT for free.
Proprietary models you can't audit, swap, or control.
Their AI runs on a model you didn't choose, trained on data you can't inspect, with costs hidden in your subscription. When EU AI Act audits come, you'll have nothing to show.
Your best AI tools have no idea what's in your hiring pipeline.
You use Claude or ChatGPT daily, but they can't see your candidates, stages, reviews, or salary bands. Two systems you already pay for, and neither can read the other.
Compare our last 3 engineering hires — time-to-fill, source, and offer acceptance rate.
Ask about your pipeline. Watch which tool it opens.
The Hiring Assistant reads your candidates, stages, reviews, and salary bands. Every answer shows the tool it called and the rows it read.
Try a question
6 agents. 160+ tools. A human still clicks the irreversible ones.
Each agent ships with its own tool list, its own instructions, and its own model tier. The Slack Bot's 10 tools are not the Hiring Assistant's 55.
Hiring Assistant
Premium TierSearch candidates, compare salary bands, draft emails, schedule interviews. 55 tools, and the transcript names each one as it runs.
Security Agent
Premium TierTriage vulnerability reports, assess severity, benchmark bounties, and draft researcher responses. Built for CSIRT teams running bug bounty programs.
Slack Bot
Economy TierMention @KitBot in any candidate channel. It reads the thread, answers questions, saves notes, and can advance or reject candidates without anyone opening Kit.
Outreach Drafter
Standard TierResearches prospects via web search, reads their profiles, and drafts personalized outreach emails. Writes complete emails, not templates.
Outreach Responder
Standard TierAnalyzes reply sentiment, reads conversation history, and drafts contextual follow-up emails. Human approval required before sending.
Email Triage
Standard TierProcesses your catch-all inbox automatically. Matches candidates to open roles, creates applications, or sends polite declines.
Choose your AI provider. Switch anytime.
Bring your own API keys from Gemini, Anthropic, or OpenRouter. Kit routes each task to a tier — economy for a lookup, premium for an analysis. Your provider bills you at its published rate and Kit takes no cut.
- Bring your own API keys, or spend Kit's starter credits
- Pay your provider directly, with no markup on tokens
- Pick the model for each tier, Premium through Embedding
- Daily and monthly credit limits per provider
- Full cost breakdown by agent, model, and token type
Run the agents on your own key, or spend Kit's starter credits.
Free for 30 days. Cancel anytime.
160+ tools, each labelled before it runs.
Connect Kit to Claude Desktop, ChatGPT, Cursor, or GitHub Copilot. Your client reads the hiring pipeline, the security program, outreach campaigns, and compensation data, and sees per tool whether running it writes anything or emails a person.
- OAuth 2.0 server with PKCE and refresh tokens
- Dynamic client registration (RFC 7591)
- Auto-approved for Claude and ChatGPT
- 8 embedding-powered semantic search indices
- Public MCP endpoint for docs and templates
OAuth 2.0 with PKCE. Scoped tokens. EU infrastructure, signed DPA on request.
The work that happens with no chat window open.
Five workflows screen reports, triage the catch-all inbox, extract resume fields, draft outreach, and read replies. The drafts wait for your approval; the rest just runs.
Inbound Email Triage
Catch-all inbox emails are automatically matched to open roles. Qualified candidates get applications created. Others get polite, personalized responses.
Outreach Email Drafting
The Outreach Drafter researches each prospect — their company, role, and recent work — then writes a fully personalized email. No templates, no merge tags.
Reply Analysis
When a prospect replies, Kit analyzes the sentiment, reads the full conversation history, and drafts a contextual follow-up. You approve before it sends.
Security Report Screening
Incoming vulnerability reports are automatically screened for 12 spam and low-quality signals. Obvious noise is flagged before your team sees it.
Resume Data Extraction
Custom data fields are extracted from resumes, form responses, and questionnaires — with confidence scores. Results broadcast to Slack candidate channels.
Kit vs. typical ATS AI features.
Most applicant tracking systems ship a job-description generator and call it AI. Row by row, here is what Kit ships instead.
| Capability | Kit | Typical ATS |
|---|---|---|
| AI model choice | BYOK: Gemini, Anthropic, OpenRouter | Locked to vendor's model |
| MCP tool access | 160+ tools for Claude, ChatGPT, Cursor | None |
| AI agents | 6 specialized agents across 3 modules | Job description generator |
| Data governance | Your keys, your data policies, full audit trail | Data sent to vendor's model |
| Cost visibility | Per-provider credits, usage dashboard | Hidden in subscription |
| AI pricing | No AI add-on line item, or bring your own key | $50-200/mo add-on |
| Autonomous workflows | 5 background automations | Manual triggers only |
| EU AI Act readiness | Your key, your model, attribution on every output | No tooling provided |
What the agents actually touch.
Six agents, 160+ MCP tools, and five workflows that run with no chat window open.
6 Specialized Agents
Hiring Assistant, Security Agent, Slack Bot, Outreach Drafter, Outreach Responder, and Email Triage. Each has domain-specific instructions and scoped tool access.
Credit Management
Set daily or monthly spending limits per provider. Threshold notifications warn you before you hit caps. Usage dashboard breaks down costs by agent and model.
Smart Model Routing
Four task tiers — Premium, Standard, Economy, Embedding — route each request to the optimal model. Fast tasks use cheap models. Complex reasoning uses premium.
Structured Outputs
Agents return fields you can sort — candidate scores, salary comparisons, severity ratings. Prose only where a human has to read it.
EU AI Act Ready
BYOK means you pick the model and can audit every decision. Irreversible actions wait for a human click. Every output is stamped with its model, timestamp, and context.
Works Without AI
If your provider goes down, the pipeline still moves. You lose the drafts and the summaries; the candidates, stages, and emails stay where they are.
Start using AI in 60 seconds.
No ML team required. No model training. Just an API key.
Choose your provider
Connect your Gemini, Anthropic, or OpenRouter API key in Settings. Or skip this — Kit includes starter credits.
Set credit limits
Define daily and monthly budgets per provider. Kit suggests sensible defaults. You can change them anytime.
Start chatting
Open the Hiring Assistant, ask a question, and watch it call the right tools. Or point Claude Desktop at Kit's MCP endpoint and drive Kit from there.
One key. Four modules.
One key drives hiring, security triage, outreach, and compensation research. Security, outreach, and compensation are add-ons billed on top of a seat, but they draw on the same key and the same credit ceiling.
AI hiring FAQ.
What is Kit?
Kit is an applicant tracking system for startups and growing teams, with 6 specialized agents and 160+ MCP tools that Claude Desktop, ChatGPT, and Cursor can drive. Bring your own Gemini, Anthropic, or OpenRouter key. Vulnerability disclosure, outreach, and compensation research are add-ons billed on top of a seat.
What providers and models does Kit support?
Kit supports three providers: Google Gemini (recommended), Anthropic (Claude), and OpenRouter (which gives you access to GPT-4o, Llama, and 200+ other models). Each provider maps to four task tiers — Premium, Standard, Economy, and Embedding — so Kit always uses the right model for the job. You can switch providers at any time.
Do I need my own API key?
No. Kit includes starter credits, so you can run the agents before you connect anything. When you move to daily use, connect your own Gemini, Anthropic, or OpenRouter key and you control the model, the spend ceiling, and where the data goes.
What data does the AI access?
Only data inside your account — candidates, applications, job postings, reviews, and uploaded documents. Never another account's rows, and never training data. MCP tools use OAuth 2.0 scoped tokens, so an external client sees only the scopes you approved.
Is Kit compliant with the EU AI Act?
Kit is built for it. You pick and audit your own model. Every output carries its model, timestamp, and context. Irreversible actions like rejection wait for a human click. You can switch the agents off per module.
How much does AI cost?
Agent usage comes out of your plan's credits, with no separate line item. Bring your own key and your provider bills you at its published rate. Routing sends simple lookups to economy models, and a daily and monthly limit stops the surprise.
What is MCP?
MCP (Model Context Protocol) is an open standard that lets clients like Claude Desktop, ChatGPT, and Cursor connect to external services. Kit exposes more than 160 MCP tools across hiring, security, outreach, and compensation. Connect Kit as an MCP server and your client can list candidates, advance applications, and check salary benchmarks.
Can I use Kit without AI?
Yes. Every screen works with the agents switched off, per module, in Settings. You lose the drafts and the summaries, and nothing in the pipeline stops. Plenty of teams start that way and switch the agents on later.
What happens if a provider goes down?
Kit's core functionality runs independently of any AI provider. If your provider is unavailable, AI features like the assistant and auto-drafting pause until recovery. You can configure a backup provider for automatic fallback.
Point your MCP client at your own pipeline.
Bring your own key or spend Kit's credits. Free for 30 days, card up front, cancel anytime. No sales call.