ZeNorm
Try it free→

Turn rough feature ideas into agent-ready specs.

Describe a feature. ZeNorm asks the hard questions. You get an agent-ready spec.

Start your first spec→See how it works
Works withClaude CodeCursorCodexAider
See it in action
ZeNorm spec editor: an answered question and a pending agent question alongside goals and constraints
The missing step

“Great specs are discovered, not drafted.”

— the thinking behind ZeNorm

Most agent failures start before code: vague intent, missing constraints, hidden edge cases. ZeNorm surfaces those gaps before they become implementation guesses.

What you said
> Add Stripe checkout with
> cart recovery and retry logic
Real intent. Missing the hard parts.
After the conversation
✓ 3 goals · 3 constraints
✓ 3 acceptance criteria
→ 3 agent tasks generated
Same idea. The hard questions answered.
How it works

Describe it. Answer the hard questions. Run it.

Describe what you want. ZeNorm asks about edge cases and constraints, writes the spec, then your agent runs it end-to-end.

  1. 01
    step 01
    Start with what you know
    Describe the feature in plain language. ZeNorm connects to your repo and reads your codebase — then starts asking the questions you didn't know to ask.
  2. 02
    step 02
    Answer the hard questions
    ZeNorm asks about edge cases, constraints, and assumptions: the details that become bugs when nobody calls them out. Your answers become goals, constraints, and acceptance criteria. A real spec, not a chat log.
  3. 03
    step 03
    Execute the whole spec
    When the spec is ready, run /zenorm <task-link> in Claude Code. The agent reads the full spec and implements every task end-to-end — constraints, decisions, and edge cases already captured.
Open-source CLI

One command runs the whole spec.

Author the spec in ZeNorm, then run /zenorm <task-link> in Claude Code. The command pulls the full spec and executes all its tasks, in order.

Run /zenorm <task-link> in Claude Code to execute the whole spec, or pull spec context into any coding agent. The spec carries the context; the agent does the implementation.

~/.claude/skills/zenorm/SKILL.md
# run a spec's tasks in Claude Code
$/zenorm <task-link>
→ loaded spec · checkout-flow
→ implementing tasks
→ branch zenorm/checkout-flow-2605271430
✓ on it
Pricing

Start free. Predictable after.

One plan for solo devs. Teams and orgs talk to us.

Starter
$0free

Try the full spec workflow during beta. No credit card.

  • Spec authoring + /zenorm execution
  • Beta usage included
  • Open-source CLI
  • Personal workspace
Try it free→
No credit card required
Most popular
Individual
$30per month

For ICs and founders shipping with agents daily.

  • $30 monthly usage · spec generation + agent runs
  • Stripe billing + self-serve portal
  • Personal workspace
  • Hard stop at limit — no surprise charges
Start free, upgrade later→
No credit card required
Team
Contact usper org

Shared specs, shared repos, pooled usage. Built around how your team ships.

  • Everything in Individual
  • Org workspaces with role-based access
  • Shared repository connections
  • Pooled usage and unified billing
  • Priority support
Talk to us→
No credit card required
Questions

Common questions

What people ask before signing up.

How is this different from just using Claude Code / Cursor?
Those are coding agents — they execute fast but only know what you tell them in the moment. ZeNorm sits before them. It interrogates your intent, grounds it in your codebase, and produces a spec that reduces what your agent has to infer.
How do I execute a spec's tasks?
In Claude Code, run /zenorm <task-link>. The slash command loads the spec context and implements the tasks inside Claude Code. You can also pull spec context into other agents like Cursor or Aider with the CLI. The spec is the contract; the agent is the executor.
Do I lose control?
No. The spec is visible, editable, and yours. You review it before anything runs. Nothing reaches your agent until you're satisfied with the spec.
What if the agent gets something wrong?
That's why the spec exists. When the output doesn't match, you can see which constraint, assumption, or edge case needs tightening — not just that something looks off. Fix the spec, then re-run with clearer context.
Why not just write a better prompt?
Prompts expire when the session ends. A spec persists across sessions, agents, and teammates. And a spec that's been interrogated — where someone asked 'what about timeouts?' and 'what happens on failure?' — produces dramatically better agent output than a prompt ever will.
How does usage work?
Each paid plan includes a generous monthly allowance covering spec generation, agent runs, and refinement passes — no per-action metering to track. Starter includes beta usage so you can see how ZeNorm fits your workflow before upgrading.

Give your agent clearer constraints.

Free to start. No credit card. Describe a feature, answer the edge-case questions, and hand the spec to your agent.

Start your first spec→See how it works