Codex and Coshell both write code with an AI agent. The difference is how many of you are in the room. Codex is one developer and one agent. Coshell is your whole team and one agent, in the same *live session*.
Coshell does not replace Codex. It is where your team uses it together.
Same OpenAI models, on your keys. One shared session instead of one lonely terminal.
Overview is in with the three stat cards and the recent orders table. Live on http://localhost:5173.
Cards stretch now. The grid uses align-items: stretch and each delta line is pinned to the bottom with margin-top: auto.
Added the toggle. 30 days aggregates by week and the axis relabels. 8 tests pass.
Delta sits on the top row now, right aligned, so the numbers line up down the column.
Token bucket middleware is in with 12 tests. Step 1 verified.
Moved the bucket to Redis with a Lua script so the check is atomic across replicas. 14 passing. Steps 2 and 3 verified.
Webhooks now share the limiter. 16 passing. Marking step 4 verified and moving to the flag rollout.
Node.js isn’t installed on this VM, so I can’t start the dev server yet. I can install it with the system package manager and then run it. Want me to?
Understood, Node 22 LTS from NodeSource.
Server is running at http://localhost:5173/, Vite dev server ready in 284ms.
Eight commits since main: the coupon model, the checkout hook, and a migration. No PR yet.
No. You have no PLAN, TODO, or roadmap file.
Your closest thing to a plan is .coshell/review/dod.md: integer cents, handled promises, no focused tests, no dead exports.
Drafted docs/plans/discounts.md with five steps. It is waiting for approval in the Plan tab.
Hi. Big Pickle here, ready when you are.
2, 3, 5, 7, 11, 13, 17, 19, 23, 29.
1
2
3
4
5
6
7
8
9
10
11
12
bun test after every step, load test before the flag movesProtect /v2 from bursty API keys without slowing normal traffic. One key should never be able to starve the others.
A token bucket per API key, 100 requests per minute, stored in Redis so all three replicas share state. Over-limit requests get a 429 with a Retry-After header. Ship behind the rate_limit_v2 flag and roll out gradually.
bun test after every step. Load test with 1k keys before the flag moves to 50%. Watch 429 rate and p99 in the dashboard for a day at 10%.
Overview is in with the three stat cards and the recent orders table. Live on http://localhost:5173.
Cards stretch now. The grid uses align-items: stretch and each delta line is pinned to the bottom with margin-top: auto.
Added the toggle. 30 days aggregates by week and the axis relabels. 8 tests pass.
Delta sits on the top row now, right aligned, so the numbers line up down the column.
Token bucket middleware is in with 12 tests. Step 1 verified.
Moved the bucket to Redis with a Lua script so the check is atomic across replicas. 14 passing. Steps 2 and 3 verified.
Webhooks now share the limiter. 16 passing. Marking step 4 verified and moving to the flag rollout.
Node.js isn’t installed on this VM, so I can’t start the dev server yet. I can install it with the system package manager and then run it. Want me to?
Understood, Node 22 LTS from NodeSource.
Server is running at http://localhost:5173/, Vite dev server ready in 284ms.
Eight commits since main: the coupon model, the checkout hook, and a migration. No PR yet.
No. You have no PLAN, TODO, or roadmap file.
Your closest thing to a plan is .coshell/review/dod.md: integer cents, handled promises, no focused tests, no dead exports.
Drafted docs/plans/discounts.md with five steps. It is waiting for approval in the Plan tab.
Hi. Big Pickle here, ready when you are.
2, 3, 5, 7, 11, 13, 17, 19, 23, 29.
1
2
3
4
5
6
7
8
9
10
11
12
The axis is not a better agent. It is the layer between the people. Codex is one developer and one agent. Coshell takes that same kind of agent and puts your whole team in one live session with it.
Codex pairs one developer with one agent in one terminal. Coshell puts several people in the same live agent session at once, on a shared filesystem, terminal, and browser. Everyone prompts the same agent and watches the same work land. It is closer to a Google Doc than to git.
With one agent per person you merge afterward, and you find the conflicts later. In a shared Coshell session there is no diverging copy to lose, and when two agents reach for the same lines of the same file, an overlap layer catches it and steers the second one before it clobbers the first.
Codex reloads context for every developer who runs it. Coshell's shared session ingests your repo once and keeps it warm, so each teammate reads from a context that is already built instead of paying to load it cold. The more context-heavy the work, the more that compounds across the team.
| Codex | Coshell | |
|---|---|---|
| What it is | An AI coding agent for one developer | A shared cloud session where a team works with one agent |
| People per session | One | Your whole team in one drive |
| Live multiplayer | One developer at a time | ✓ Same session, live presence and follow-along |
| Overlap detection | Merge it yourself, later | ✓ Built in, when edits hit the same lines |
| Codebase context | Reloaded per developer | ✓ Loaded once, kept warm, shared by the team |
| Models | OpenAI models | 70+ models, including OpenAI, on your keys |
| Per-session privacy | ✓ Private, named teammates, or whole drive, per session | |
| Works from | Your machine | Any browser, or your terminal over the CLI |
| Git | Your repos | Pushes to your real repos, normal git |
| Runs on your own machine | ✓ Free, self-hosted | |
| Pricing unit | Per the model usage you run | Free to self-host, no seats |
| Best for | One developer moving fast, alone | A team that wants to build with the agent together |
Nothing about your models or your git changes. Coshell is a fork of OpenCode that runs OpenAI's models on your own keys, alongside 70+ others. What is new is the layer Coshell adds: one live session your whole team shares, presence, follow-along, and overlap detection. Pay free to self-host, no seats.
Codex and Claude Code make one developer faster, alone: one person, one agent, one terminal. Coshell is the multiplayer layer. Several people work in the same live agent session at once, on a shared filesystem, terminal, and browser, with presence, follow-along, and an overlap layer that catches overlapping edits before they clobber. It is the same kind of agent, built for a team instead of a seat.
Yes. Coshell runs 70+ models, including OpenAI's, scoped per drive and billed by your provider on your own keys. Admins pick the allowed set and a fallback. You are not switching models, you are putting your team in one session that uses them together.
No. Two things prevent it. Because everyone works in one shared environment, there is never a second copy drifting out of sync, so the conflicts you would normally merge later never appear. And when two agents reach for the same lines of the same file, the overlap layer catches it and holds the second edit before it clobbers, steering that agent to take other lines, wait, or let the first land first.
No. Work in a drive commits and pushes to your own repositories like any other git remote, so your branches, reviews, and CI keep working as they do today, and nothing is locked inside Coshell. On the Enterprise tier adds org-wide on-prem and SSO; either way, self-host keeps your code on your own machine.
Keep the OpenAI models you already use. Add the room everyone codes in together. Open a drive in the browser, or attach from the CLI.