---
title: "Discussion on \"Running Claude, Codex & Grok as One Team on a Locked-Down Enterprise Project (Day 0)\" | Hashnode"
description: "Discussion on \"Running Claude, Codex & Grok as One Team on a Locked-Down Enterprise Project (Day 0)\". A 2-week experiment on a locked-down client project: can a small, measured AGENTS.md cut token cost and agent mistakes? Baseline: 98% cache reads."
url: "https://hashnode.com/posts/running-claude-codex-grok-as-one-team-on-a-locked-down-enterprise-project-day-0/6af3c09ce3dfbf696f9944dc"
---

[Feed](https://hashnode.com/feed) /

Discussion

# [Running Claude, Codex & Grok as One Team on a Locked-Down Enterprise Project (Day 0)](https://seulki-i.hashnode.dev/running-claude-codex-grok-as-one-team-on-a-locked-down-enterprise-project-day-0?utm_source=hashnode&utm_medium=feed)

A 2-week experiment on a locked-down client project: can a small, measured AGENTS.md cut token cost and agent mistakes? Baseline: 98% cache reads.

23h ago · 5 min read · seulki-i.hashnode.dev  
Read the article

-   [#ai](https://hashnode.com/tag/ai)
-   [#claude](https://hashnode.com/tag/claude)
-   [#codex](https://hashnode.com/tag/codex)
-   [#grok-bot](https://hashnode.com/tag/grok-bot)
-   [#productivity](https://hashnode.com/tag/productivity)
-   [#developer-tools](https://hashnode.com/tag/developer-tools)

## Written by

[CHOI SEULKI](https://hashnode.com/@seulki-i)

@seulki-i · Seoul, South Korea

  
[View profile](https://hashnode.com/@seulki-i)

Backend dev in Korea · Running Claude, Codex & Grok as a team in enterprise SI projects

## Responses 2

Markdown supported · ⌘↵ to post

[Ascendra Ventures](https://hashnode.com/@ascendra-ventures)

[21h ago](https://hashnode.com/posts/running-claude-codex-grok-as-one-team-on-a-locked-down-enterprise-project-day-0/6af3c09ce3dfbf696f9944dc/comment/6af3d5dfd6b5ade573041bff)

The rule that an instruction can only change if it cites a line in agent-misses.md or a measurement is the part I would copy. Instruction files tend to grow every time someone reads a new tip, and nobody removes anything. A one-page limit with add one, remove one keeps it honest.

The numbers make the case well. 98% cache reads says the agents are mostly re-reading long context, and nearly a quarter of tool calls being searches suggests they are rediscovering facts that were already written down. Mapping short requests to a fixed procedure, design doc, approval, implement, then review by a different agent, should cut both.

At Ascendra Ventures we work on client systems with similar limits, read-only repos and no direct database, and the separation between the agent that builds and the one that verifies has been the most reliable habit we have. A mapping table where every row carries its evidence file is a nice touch for legacy migration.

For the Nov 25 results, will you also track how many "done" reports the verifier rejected?

[PromptAlo](https://hashnode.com/@promptalo)

[8h ago](https://hashnode.com/posts/running-claude-codex-grok-as-one-team-on-a-locked-down-enterprise-project-day-0/6af3c09ce3dfbf696f9944dc/comment/6af3d5dfd6b5ade573041bff#reply-6af48d28c87d8460559560ae)

I would track those rejected done reports. I would also count the times an agent declared done before the verification step in the procedure had run, because those are process misses even when the verifier later catches them.