FeaturesAISecurityIntegrationsSelf-hostingDevelopersPricingUse casesWhy HarlyDocs
Live demoGet started

AI for the busywork of hiring.

Score candidates, draft emails and prep interviews with the AI provider you choose. The assistant asks before it changes anything.

Ask Harly AI. An assistant that knows your pipeline and asks before it changes anything.

  • Answers from your dataWho needs review, which jobs are at risk, what happened this week.
  • Acts when you confirmMoving a candidate, scheduling an interview or sending an email waits for your click.
  • UndoReverse stage moves, new tasks and scheduled interviews.
  • Your permissionsThe assistant can only do what you can do, checked again on every confirm.
  • Automations in chatDescribe a workflow. Harly AI drafts it, simulates it and applies it after you approve.

Score candidates, with the evidence. Every score shows the lines from the résumé that support it.

  • Scoring with quotesEach criterion links to a quote, and quotes are checked against the résumé.
  • Rules firstA built-in rules engine scores every candidate. AI adds to it, and Harly falls back to rules if the model fails.
  • Score on applyScore new applications as they arrive, one by one or in bulk.
  • Talent pool matchesFind the five closest past candidates for a new job. Needs an OpenAI key.
  • Duplicate detectionSpot the same person applying twice.
  • Résumé parsingFill in the application and build a structured profile from a CV.

Less writing. Better prep. Drafts and summaries where you already work.

  • Job descriptionsDrafted from a short brief, in the voice of your career page.
  • Screening questionsSuggested questions for the application form.
  • EmailsDraft candidate emails, summarize a thread and suggest a reply.
  • Interview briefsA summary, what to probe, questions to ask and red flags.
  • Interview notesTurn rough notes into a clear summary.
  • ScorecardsSuggested attributes and clearer feedback comments.

You decide what the model sees. AI is off until you turn it on, and you can hide candidate identity from the model.

  • Off by defaultAI features stay hidden until you add a key. Auto-scoring and duplicate checks have their own switches.
  • Resume anonymizationHide names and contact details from your provider. Features that need a name pause while it’s on.
  • Encrypted keysYour API key is encrypted with AES-256-GCM and never sent back to the browser.
  • Prompt-injection guardrailsBuilt-in protection against instructions hidden in résumés and emails.
  • Short memoryChats are deleted after 90 days, and when a candidate is erased.

Bring your own model. Fourteen providers, local models, or any OpenAI-compatible endpoint. Pick a model from the catalog or type its ID.

OpenAI
Anthropic
Google Gemini
xAI
OpenRouter
Mistral
DeepSeek
Groq
Together AI
Fireworks
Cerebras
Azure OpenAI
Ollama
OpenAI-compatible

Works without AI, too. Turn AI off and Harly keeps working.

Rules-based scoring

A deterministic engine with relaxed, balanced and strict modes.

Résumé autofill

Heuristics fill in the application form from an uploaded CV.

No AI clutter

AI buttons, panels and the chat launcher stay hidden until you turn AI on.

Questions.

No. It scores, drafts and summarizes. Every change Harly AI proposes waits for a person to confirm it. Automations can use a score in a condition, but only rules your team writes, and there is no “AI decides” action.

OpenAI, Anthropic, Google Gemini, xAI, OpenRouter, Mistral, DeepSeek, Groq, Together AI, Fireworks, Cerebras, Azure OpenAI, Ollama, or any OpenAI-compatible endpoint. Talent pool matching needs an OpenAI key.

Yes. Use Ollama or any OpenAI-compatible server such as vLLM or LM Studio. Local addresses are blocked by default; set HARLY_ALLOW_LOCAL_LLM=1 to allow them.

Harly doesn’t charge for AI. You pay your provider directly for what you use.

Yes. Scoring uses the built-in rules engine, résumé parsing uses heuristics, and AI buttons stay hidden until you turn AI on.

Hiring software, on your terms.

Open source. Self-hosted. Built for growing teams.