Agents aren’t following your code standards

Every agent reads its own instruction file, so your standards get copied into every repo and drift apart with every update.

proto[code] serves one library of your team’s code standards to every agent, before it writes.

None of your agents share instructions

Your rules get copied into every repo, for every agent, and every copy drifts as updates are made.

The files proto[code] replaces: CLAUDE.md, AGENTS.md, .cursor/rules, copilot-instructions.md, GEMINI.md, .windsurf/rules.

Agents look it up before they write

Give an agent a task. Before it writes a line, it asks proto[code] which of your standards apply, then writes to them.

Different agents define tasks in different ways. We accommodate and provide consistent responses every time.

Task for Claude Code, in shipment-api

Add an async repository method

proto[code] returned 3 standards

  1. Async/Await Method Naming ConventionMethods that return a Task end in Async.
  2. Await or Return Inner TaskAlways await or return the Task from Task.Run.
  3. Never Return Void from AsyncAsync methods return a Task, never void.

Task for Cursor, in tracking-service

Add an endpoint that returns a driver’s name

proto[code] returned 3 standards

  1. APIs Must Return JSON ObjectsReturn an object, never a bare string or number.
  2. Singular Resource Segments in API Routes/api/user/{id}, never /api/users/{id}.
  3. No Logic in ControllersControllers hand off; the logic lives elsewhere.

Task for Codex, in billing-service

Save a new invoice record

proto[code] returned 3 standards

  1. Use Value Generators for EF-Generated ValuesKeys and timestamps come from value generators.
  2. Entity Framework Model Naming and EncapsulationEF models end in Record and stay in the repository.
  3. No In-line SQLQueries go through LINQ, never SQL strings.

How a rule travels

1. Write it once

# Async methods end in Async.

C#.NET

2. Scope it to an agent, a repo, or everything

Each rule applies to some agents and some repos. The brackets show which.

3. Every agent reads it

  • Claude Code, on its next task
  • Cursor, on its next task
  • Codex, on its next task
  • Copilot, on its next task
  • Gemini CLI, on its next task

See what your agents read

Every lookup is counted. Find the standards your agents reach for most, and the ones that have gone stale.

How often agents read each standard, week by week, over the last 12 weeks
12 weeks ago11 weeks ago10 weeks ago9 weeks ago8 weeks ago7 weeks ago6 weeks ago5 weeks ago4 weeks ago3 weeks agoLast weekThis week
Async/Await Method Naming ConventionMost readRead oftenRead mostRead oftenRead mostRead oftenRead most
APIs Must Return JSON ObjectsRead oftenRead mostRead oftenRead mostRead oftenRead mostRead oftenRead most
Singular Resource Segments in API RoutesRead sometimesRead oftenRead mostRead oftenRead mostRead oftenRead most
No Logic in ControllersRead sometimesRead oftenRead sometimesRead oftenRead sometimesRead oftenRead sometimesRead often
No In-line SQLNew, and climbingNot readRead rarelyRead sometimesRead oftenRead most
Use Value Generators for EF-Generated ValuesRead rarelyRead sometimesRead rarelyRead sometimesRead rarelyRead sometimesRead rarelyRead sometimesRead rarelyRead sometimes
Avoid Compound StatementsNo reads in 6 weeksRead sometimesRead rarelyNot read
Python Variable Prefix RequirementNo reads in 10 weeksRead rarelyNot read
Not readFewer readsMore

We built it because we needed it

Our own team works from it every day, on our own code.

“Before using it, every AI-generated change felt a little different.

Now our agents work from the same playbook, so onboarding new projects is faster, reviews are easier, and the quality is much more consistent across the team.”

MattApplication Architect

“Managing AI instructions across 15 applications used to be a configuration nightmare.

Now we update our agent instructions in one place and know every agent is always up to date.”

LaurenTeam Lead

“As someone jumping between multiple applications, it’s a huge relief.

I know the agent already understands our standards and workflows because they’re managed centrally instead of being duplicated across every codebase.”

JamesSoftware Engineer

Pricing

Try it free for 14 days. You only pay once you’ve seen it work.

What each plan includes
Free$0 / monthPro$9 / monthor $90 a yearTeam$14 per user / monthor $140 per user a yearEnterpriseContact us
Users112–20Unlimited
Code standards252501,000Unlimited
Agent instructions1050250Unlimited
Connected agents2UnlimitedUnlimitedUnlimited
Lookups per month1,000UnlimitedUnlimitedUnlimited
Document storage50 MB1 GB10 GBUnlimited
GitHubIncludedIncludedIncludedIncluded
Azure DevOpsNot includedIncludedIncludedIncluded
Agent analyticsNot includedIncludedIncludedIncluded
AI-drafted standards from agent feedbackNot includedIncludedIncludedIncluded
Shared team libraryNot includedNot includedIncludedIncluded
SSO and RBACNot includedNot includedNot includedIncluded
StartStart with ProStart with TeamTalk to us

You’re one command away from better code

Pick your agent and copy one line. It signs in the first time it connects.

Claude Code

Claude Code

reads CLAUDE.md

Replaces CLAUDE.md in every repo.

claude mcp add --transport http protocode https://mcp.code.prototypd.dev

Cursor

reads .cursor/rules

Replaces .cursor/rules in every repo.

{
  "mcpServers": {
    "protocode": {
      "type": "http",
      "url": "https://mcp.code.prototypd.dev"
    }
  }
}

Gemini CLI

reads GEMINI.md

Replaces GEMINI.md in every repo.

gemini mcp add --transport http protocode https://mcp.code.prototypd.dev

GitHub Copilot

reads copilot-instructions.md

Replaces copilot-instructions.md in every repo.

{
  "mcpServers": {
    "protocode": {
      "type": "http",
      "url": "https://mcp.code.prototypd.dev"
    }
  }
}

Codex and others

reads AGENTS.md

Any agent that speaks MCP uses the same address.

"mcpServers": {
  "protocode": {
    "type": "http",
    "url": "https://mcp.code.prototypd.dev"
  }
}

FAQ

How does an agent know which standards apply?

It describes the task it’s about to do. proto[code] returns only the standards that match that task, so the agent’s context holds the rules it needs and nothing else.

What happens when I change a rule?

Every agent gets the new version on its next task. Nothing to commit, and no pull requests to open across your repos.

Can a rule apply to only some repos or agents?

Yes. Scope each rule to an agent, a repo, or everything. A rule for one service never reaches an agent working on another.

What’s the difference between agent instructions and code standards?

Agent instructions shape how an agent works: your workflow, your tools, how it reports back. Code standards are rules for the code it writes, served only when a task needs them.

How do I find rules nobody uses?

Every lookup is counted. Agent analytics show the standards your agents read most and the ones they haven’t read in weeks, so you can retire them.