Claude Code and Codex made each engineer faster, in their own session. Coshell is the layer that puts the whole team and the agent in one shared, live session, across more than 70 models.
Coshell is not a better single-player agent than Claude Code.
It works with it, and across more than 70 other models, to make the whole team work with the agent together.
Claude Code is an excellent agent for one person in one session. The moment a team works on the same codebase, that single-player model creates friction: duplicated context, work done in isolation nobody else can see, and no shared way of operating. Coshell does not replace the agent. It is the collaboration layer those tools do not have.
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
| Claude Code | Coshell | |
|---|---|---|
| What it is | A single-player AI coding agent for one engineer in one session | A multiplayer cloud session where a whole team prompts one live agent together |
| Session model | One private session per person | One shared live session several people work in at once |
| Real-time multiplayer presence | Sessions are individual | ✓ See who is in the session, follow along, prompt the same agent |
| Overlap detection | Parallel work diverges and is merged later in git | ✓ Catches overlapping edits before they clobber |
| Load the codebase once | Each session reloads context; every developer pays to load it again | ✓ A shared warm session loads context once and keeps it cached |
| Per-session privacy | ✓ Private to you, shared with teammates, or open to the drive | |
| Models | Anthropic models | 70+ models across providers, on your own keys |
| Works from | Local terminal | Browser, CLI, or phone. The session lives in the cloud |
| Pushes to real git | ||
| Best for | One engineer moving fast in their own session | A team that wants to work with the agent together on one codebase |
Coshell works with those agents rather than replacing them, and across more than 70 models. The single-player tools made each individual faster while leaving the team's coordination worse. Coshell is the layer that makes the team work with the agent together, through shared live sessions, overlap handling when code is written, shared context, and a standardized environment. It is the collaboration layer those tools do not have.
Those features run several agents, or give each teammate their own session. Coshell is the opposite shape: several people in one shared live session, prompting the same agent on the same running environment, with an overlap layer when code is written. It is humans collaborating with one agent, not many agents working alone. Read why multiplayer is the axis.
The core of Coshell is a coordination layer that works across agents and across vendors, including different model providers in the same session. A model provider is structurally disincentivized to build that, because it requires being model-agnostic, which works against their own model. Their job is to make their agent better on its own. The neutral layer that lets any agent coordinate with any other is something only a party that does not own a model is positioned to build, and it compounds on coordination data over time.
Usually less. A shared, long-running session ingests your codebase context once and keeps it warm, where a team in separate sessions each pay to load that same context over and over. Cached context is read at a fraction of the price of fresh input, so you pay to load the codebase once for the team rather than once per developer. Coshell is free to self-host and covers your whole team, so adding people is not penalized.
Open a drive in the browser, or attach from the CLI. Your whole team works in shared live sessions with the agent, on the same codebase. Need your own infrastructure? Coshell runs on yours on the Enterprise tier.