Org Brain · Just shipped
Ask your org.
Get the answer.
You already have the data. Your frameworks, goals, evaluations, and every 1:1 log all sit in Harmny. Your Org Brain reads them and tells you the answer when you ask: who’s ready for the next level, what stalled this week, where the team needs you next.
Engineering framework · 14 evaluations · 86 1:1 logs · 23 active goals
Three engineers are above 80% on your L5 rubric:
- AS
Ana Silva
4 of 5Missing: cross-team initiative as lead
- MW
Marcus Webb
4 of 5Missing: signed-off system design review
- LC
Lin Chen
3 of 5On track in current development plan
Want me to draft promotion-readiness notes for any of them, or pull their open goals?
Sarah (CEO) asked: "Where are goals stalling?" · 4s ago
Running on gut feel?
You have the data. You don't have the answer.
Tuesday · board prep
"How is the org actually doing?"
The data is in five tools. The answer is in nobody's head. You either guess, or you ask three managers to spend a week pulling it together — and still get an opinion.
Quarterly planning
"Who's ready to lead next?"
You suspect three names. The framework, evaluations, and 1:1 history that would confirm it sit in three different views. So you promote on memory, not signal — and find out you missed someone two quarters later.
Review cycle week
HR loses three weeks compiling what should be a button click.
Manager notes, ratings, goal progress, calibration data — all yours, all scattered. Every cycle, the same compile job. Every cycle, the same silent tax on the people who should be doing strategy.
You're not missing data. You're missing a way to ask it. Your Org Brain reads what's already there and answers in a sentence, so you can decide instead of dig.
The pattern underneath
Motion isn’t progress.
Everyone’s calendar is full. Slack is loud. Reviews keep slipping into next cycle. 1:1s happen every week and nothing actually shifts. The team isn’t slacking. They’re swamped. The work just isn’t moving.
You can’t see the pattern because the pattern isn’t in any one place. It lives in what’s not happening: the goals quietly untouched, the team whose 1:1 frequency dropped before their delivery did, the cycle that missed twice in a row with no flag in the tracker. Five managers will give you five different stories. None of them are wrong. None of them connect.
So you do what executives do when the picture won’t come into focus. You guess. You delegate. You spend a week running the compile yourself. None of those are leadership. They’re tax.
“Which goals haven’t been touched in 30 days?”
6 goals across Engineering and Marketing. Sarah owns 3 of them. None have been updated since Q3 OKR setting.
“Where are 1:1s happening but goals aren’t moving?”
Engineering team B. Meetings every week, goal progress flat for two cycles. The conversation isn’t translating into action.
“Which managers haven’t run a single review this cycle?”
Three managers: Mike, Priya, Jordan. Deadline was 8 days ago. Their 11 reports are blocked on you.
The chaos lives in the gap between what your team is doing and the picture you have of it. Close that gap. Busy and progress stop looking the same once you can actually tell them apart.
In-workflow AI
One button. The draft you'd have written. Ten seconds.
AI lives where the work already happens. On the employee profile, a "Suggest development plan" button reads their competency scores, their last evaluation, and their goals — then drafts a 90-day plan that fits your framework, not a generic template. In a review cycle, "Draft summary" turns a year of manager notes and ratings into a narrative ready to publish. Before a 1:1, "Prepare talking points" pulls open goals, recent feedback, and last meeting's action items into a list you can skim in 30 seconds.
- Suggest development plan — uses your competency framework and the employee's actual gaps
- Draft cycle summary — synthesizes notes, ratings, and goal progress into a publish-ready narrative
- Prepare 1:1 talking points — open goals, recent feedback, last meeting's action items
- Every output opens in your normal editor — review, edit, save. Or discard.
Ana Silva
Senior Engineer · L4 → L5
Last review
4.2
Goals
3 open
Gap
Stakeholder mgmt
90-day plan: Stakeholder management for L4 → L5
- · Lead a cross-team initiative — Q3 migration is open and unowned
- · Run two stakeholder reviews — using the framework's "negotiation" rubric
- · Capture decisions in writing — pattern from your existing playbook
Always-on chat
A chat that knows your framework, your people, your last six 1:1s.
ChatGPT will give you a generic answer. Harmny will tell you what your L5 looks like, who's closest to it, and which goals they need to close. The chat widget lives in the bottom-right of every page, scoped to your org's data only. Ask in plain language. Get an answer grounded in what you actually have.
You
"Which engineers are closest to senior?"
Harmny Intelligence
Three people are above 80% on the L5 rubric: Ana Silva (4 of 5 criteria, missing cross-team lead), Marcus Webb (4 of 5, missing system design review), and Lin Chen (3 of 5, on track in their current plan). I can draft promotion-readiness notes for any of them.
Harmny Intelligence
Scoped to your orgBring your own model
Your model. Your contract. Your audit log.
Drop in an Anthropic, OpenAI, or compatible API key in Settings. Harmny encrypts it at rest with AES-256-GCM and routes every Intelligence call through your account. You inherit the contract you already signed — DPA, region, data-retention terms, the rate you negotiated. Token cost shows up on your provider bill instead of Harmny's, and you can revoke the key any time without uninstalling.
- Anthropic, OpenAI, or compatible — switch providers without changing how you work in Harmny
- AES-256-GCM at rest — keys never leave the database in plaintext, never returned to the browser
- SSRF-hardened endpoint config — custom base URLs are validated, private addresses rejected
- No metering when you BYO — usage runs against your provider account, not Harmny's pool
Settings · Intelligence · Provider
MCP server
Connect Harmny to Claude Desktop. Or your own agent.
Generate an API key, paste two lines into your Claude Desktop config, and your AI assistant can read your frameworks, employees, goals, tasks, and projects in any chat — with the same scoping rules Harmny enforces internally. It's the open MCP standard, so the same key works with Cursor, with custom agents, with anything that speaks the protocol.
- Per-key scopes (read-only by default) and per-key rate limits — least privilege, enforced
- Bearer token auth, rotation in one click, full audit log per tool invocation
- Org-scoped at the database layer — same multi-tenant guards as the in-app product
{
"mcpServers": {
"harmny": {
"command": "npx",
"args": ["-y", "mcp-remote", "https://harmny.ai/api/mcp"],
"env": {
"MCP_AUTH_TOKEN": "hmn_•••••••••••••••••"
}
}
}
}get_frameworks
Career ladders & competencies
get_employees
Roles, levels, managers
get_goals
Open goals + progress
get_tasks
Tasks & projects
Why this isn't another bolt-on
Most HR AI is a chat layer over a legacy schema.
That's why the answers feel generic. The underlying data model was never built for an AI to reason about — competencies are tags, frameworks are PDFs, goals are spreadsheets. So the AI hallucinates the structure that should already be there.
Harmny started structured. Frameworks, levels, competencies, evaluations, ratings — first-class objects with relationships, not free text fields. Intelligence isn't bolted on. It's the natural expression of a data model that was already AI-shaped.
That's the moat competitors can't ship in a quarter.
Privacy by design
Your data stays your data.
Every concern an enterprise security team will raise, answered in the product before they ask.
No training on your data
Harmny doesn't train models on your content. Anthropic and OpenAI don't train on API traffic by default. Your prompts answer your questions and stop.
Multi-tenant isolation
Every query is scoped to your organization at the database layer. No cross-tenant context, no shared embeddings, no exceptions.
AES-256-GCM at rest
BYO API keys are encrypted with authenticated AES-256-GCM and never returned to the browser. Rotation in one click.
Human gate, every output
No AI artifact persists without an explicit save click. Drafts are draftable, editable, discardable. The human decides what counts.
Bring your own key
If your security team won't let model traffic flow through Harmny's account, route it through yours. Your DPA, your region, your audit log.
EU AI Act safe by design
Outputs framed as factual signals, not employment decisions. Team-level patterns surface before individual ones. Human-in-the-loop is structural, not policy.
"The 'Suggest development plan' button gave me a draft that knew our framework better than I did. I edited two bullets and saved. That used to be a full afternoon."
Maya Patel
Engineering Manager, Stackform
What teams ship in their first week with Intelligence
Review packets that draft themselves
A year of manager notes plus competency scores plus goal progress, distilled into a narrative ready to publish. Edit once, save, move to the next direct report.
1:1s that don't start cold
Open the meeting. Click "Prepare talking points". Walk in with the open goals, the recent feedback, and the action items from last time already in front of you.
Dev plans your framework actually backs
Generic templates produce generic plans. Yours uses your competency definitions, your level criteria, your existing goals. The plan reads like the framework, because it is.
Privacy your security team will sign off on
Org-scoped queries, encrypted credentials, optional BYO key, no model training. The hard questions answered in the product, not in a deck.
AI that follows you out of Harmny
MCP turns Claude Desktop, Cursor, or any compatible agent into a Harmny-aware assistant. Generate a key, paste a config, your data is one prompt away — wherever you work.
Friday afternoon back
The blank page is the bottleneck. Intelligence removes it. The 45-minute task takes 10 seconds, you spend 2 minutes editing, and you go home on time.
Included, not upsold
On Growth and above. No separate AI SKU.
Each plan includes a monthly token allotment for Intelligence and MCP. Run past it and the rest of Harmny keeps working — or drop in your own API key for unmetered usage. No per-seat AI fee, no usage-based billing surprise at the end of the quarter.
Intelligence — what teams ask before turning it on
Privacy, providers, MCP, pricing — the questions security and operations leads keep asking.
Explore more features
Core HR
Performance Reviews
Run fair, structured evaluation cycles without the chaos.
Learn moreCore HR
Competency Frameworks
Define clear skill levels and career ladders for every role.
Learn moreCore HR
Goals & OKRs
Align company, team, and individual objectives in one view.
Learn moreCore HR
1:1 Meetings
Shared agendas, action items, and conversation history.
Learn moreCore HR
360 Feedback
Collect peer feedback to build a complete picture.
Learn moreCore HR
Development Plans
Personalized growth paths with trackable milestones.
Learn moreClarity. Motivation. Growth. For your whole team.
Free forever for up to 10 users. No credit card required. Full access from day one.
By signing up you agree to our Terms of Service and Privacy Policy.