Integration

Kriyastream + Claude Code

Claude Code is exceptional at implementing software in a repo. Kriyastream is exceptional at holding product meaning across sessions. Together: Claude Code ships Features; the Product Map keeps those Features coherent for the next human or agent who opens the project.

What each tool owns

Claude Code owns implementation: editing files, running commands, iterating toward a working change. It is repo- and session-aware, but product lore still collapses when the conversation ends or a new teammate starts fresh.

Kriyastream owns product structure: Domain → JTBD → Experience → Capability → Feature, plus gaps and code evidence. That is the durable brief Claude Code should inherit — not a paragraph you retype every Monday.

  • Claude Code = how we change the codebase
  • Kriyastream = what the product is and what is missing
  • Together = agent speed without product amnesia

The failure mode this pairing fixes

Solo vibe coding with Claude Code often produces a demo that works and a product nobody can explain. The next session invents a second auth flow, skips billing, or “improves” a Feature that was never named.

Issue trackers do not fix that by themselves — they store work items, not product completeness. A Product Map makes absence visible and gives Claude Code a Feature-sized target.

Recommended operating loop

Keep the Product Map current in Kriyastream (from-code inference plus human review). When you open Claude Code, scope the session to one Feature or Capability: name, description, domain, and known paths.

After the PR merges, update Feature evidence and glance at related gaps. The next Claude Code session starts from a stronger map instead of a colder chat log.

  • Map first — or infer from the repo, then correct names
  • One Feature per focused session when possible
  • PR description cites the Feature (not only a ticket)
  • Refresh code links when paths change

CLAUDE.md / rules vs Product Map

Project rules tell Claude Code how to behave in this repository (style, commands, constraints). A Product Map tells it what the product is. You want both.

Putting your entire product encyclopedia into a rules file creates brittle, unreadably long instructions that drift. Put product structure in Kriyastream; keep rules lean and behavioral.

Example session brief

Weak: “Improve checkout.” Strong: “Implement Feature: Apply discount codes (Domain: Storefront Shopping). Respect existing cart capability; do not invent a second promotions engine. Paths: … Map: [link].”

That brief is copy-pasteable into Claude Code and reusable by a human reviewer — because it is grounded in the map, not improvised.

Honest boundaries

Kriyastream is not a Claude Code plugin that auto-drives the terminal. It is the product intelligence layer your agents and humans share. Claude Code remains the implementation surface.

If you only use Claude Code without a map, you still get speed. You do not get durable product context across people, tools, and time.

How to do it

  1. 1Create or infer a Product Map for the repo Claude Code will edit.
  2. 2Pick one Feature (or small Capability) as the session target; copy name, intent, and paths.
  3. 3Run Claude Code against that brief; keep scope from expanding into unnamed Features.
  4. 4Open a PR that cites the Feature; merge; update map evidence and gaps.
  5. 5Start the next session from the updated map — not from memory of the last chat.

FAQ

Does Kriyastream install inside Claude Code?

No install-inside requirement. Kriyastream is the durable product layer you keep alongside Claude Code. Use the Product Map as the source of session briefs and provenance.

Can I just paste the whole Product Map into the prompt?

You can paste a slice (domain + target Feature). Dumping the entire map every time wastes context. Prefer Feature-scoped briefs with links back to the live map.

How is this different from Cursor?

Same product layer, different IDE surface. See Kriyastream + Cursor. Many teams use both tools against one Product Map.

Where do public Product Maps fit?

Use them as structural references for products like yours while you build your own from-code map — especially useful before your repo has rich coverage.