AI coding agents in Slack, compared: plans and billing
Every AI coding agent you can @mention in Slack, compared: which plans include it, what a mention bills, DMs, stop controls and what comes back.
Nearly every AI coding agent answers a Slack mention now. Anthropic, OpenAI, Cursor, Cognition, GitHub, Factory and Linear all ship one, and Slack has added code channels built for agents, so having a Slack app doesn't separate one tool from another any more. If you're picking which bot your team gets, look at the bill first. Each mention needs some plan and draws on some meter, and what comes back ranges from a chat answer to a link to a web session to a pull request.
We build Sinatra, which is one of the agents in the table, so read this as a vendor writing about its competitors. Sinatra has several gaps, and they're in the table with everyone else's.
Product details checked October 2026 against each vendor's own docs and pricing pages, linked throughout. Several of these integrations shipped or changed in the last four months; check the vendor pages for current specifics.
Slack coding agents at a glance
| Agent | You mention | DMs | Stop from Slack | Teams app | Plans that include it | What a mention bills | What comes back |
|---|---|---|---|---|---|---|---|
| Claude Tag | @Claude | Yes | Yes | Waitlist | Team, Enterprise | Org usage balance | Answer, doc, or draft PR |
| Claude Code in Slack (older app) | @Claude | No | Not documented | Not documented | Pro, Max | Your plan's limits | Link to a cloud session, Create PR button |
| ChatGPT (Codex) | @ChatGPT | Yes | When the Cancel control is shown | Yes | Business, Edu, Enterprise (pricing page also lists Plus) | Workspace credits | Result in the thread |
| Cursor | @cursor | Not documented | Yes | Yes | Premium ($30) and up, usage-based pricing on | Model API pricing | PR link |
| Devin | @Devin | Not documented | Yes, by keyword | Yes | Pro, Max, Teams, Enterprise | Plan allowance, then usage at API pricing | Replies in the thread, PR |
| GitHub Copilot | @GitHub | Yes | In code channels | Yes | Business, Enterprise (public preview) | AI credits | PR link |
| Factory | @Factory | Yes | Not documented | Yes | Teams and up | Not stated on the Slack page | Response, linked PR |
| Linear | @Linear | Group DMs, not 1:1 | Not documented | Linear Agent, yes | Basic, Business, Enterprise | Prepaid dollar balance | Draft PR |
| Sinatra | @Sinatra | No | No | No | Every plan, Free included | Nothing per mention | Answer in the thread, PR linked |
"Not documented" means we couldn't find it on the vendor's own pages, not that the feature is missing. Treat those cells as questions for the vendor.
What a Slack mention costs on each
Claude Tag
Anthropic's Claude Tag bills by usage against the organization. Its overview says adding Claude to Slack "doesn't add a per-seat charge," and that channel, thread and group DM work "draws from a usage balance." On Team plans that balance has to be funded before anything runs, an Owner sets a spend limit, and Anthropic says "there's no published per-task cost guidance." One-to-one DMs are different: once the sender has connected a Claude account, they run on that account and seat. Until then, Anthropic says a DM "can bill to your organization for a limited time."
Claude Code in Slack, the older app
The older Claude Code in Slack app bills the person. Anthropic's Slack docs list Pro or Max as the plan and say "Each session runs under your own Claude account, using your connected repositories and your plan limits." Since October 5, 2026 it no longer answers in channels of Claude Tag-paired workspaces, so for most companies on Team or Enterprise this has become Claude Tag.
ChatGPT, formerly @Codex
ChatGPT's Slack app is the renamed Codex bot. OpenAI's Slack page says "If you previously used @Codex, start new requests with @ChatGPT after your admin enables the ChatGPT app," and that ChatGPT can hand repository work to a separate Codex cloud task when the workspace has cloud delegation on. The admin article says "Your workspace needs credits to use @ChatGPT," and that work a member authorizes with their personal connections runs under their account. OpenAI's help center lists Business, Edu and Enterprise. Its pricing page also lists a "Slack integration" under Plus, so if you're on Plus, check with OpenAI which one applies to you.
Cursor
Cursor bills model tokens at API prices. Setup requires that you "enable usage-based pricing," per its Slack integration docs, and its cloud agent docs say "Cloud Agents are charged at API pricing for the selected model," with a spend limit you set. Its pricing page lists cloud agents on Premium, at $30 a month, and the plans above it.
Devin
Devin includes Slack on its paid plans. The pricing page lists Pro at $20 a month, Max at $200 a month, and Teams at "$80 /month for team plan + $40/mo per full dev seat," an $80 minimum that covers two full seats, and its comparison table marks Slack and Teams as included on everything except Free. "Each paid plan comes with a usage allowance that refreshes automatically on a daily and weekly basis," and past that you buy extra usage "consumed at API pricing." Enterprise is billed in ACUs.
GitHub Copilot
Copilot bills AI credits. GitHub's August changelog says "Usage counts against your existing Copilot entitlements and can be managed with existing Copilot cloud agent budgets." That changelog makes the new experience a public preview for organizations on Copilot Business and Enterprise, while GitHub's setup docs say "Available for all paid Copilot plans," and we couldn't find a GitHub page that reconciles the two for Pro and Pro+. Our Copilot pricing post explains how fast AI credits go on agent work.
Factory
Factory needs "a Factory organization on a Teams plan or higher," per its Slack docs, and its pricing puts Teams at "$60/mo + $40/seat," for up to 10 seats. The Slack page doesn't say what a single mention draws down.
Linear
Linear bills coding sessions from a prepaid workspace balance in dollars. Its AI credits page says each session costs "model tokens at provider-published rates, with no markup, and sandbox runtime at $0.25 per 20-minute block," on Basic, Business and Enterprise. A coding session can start from a Slack thread with @Linear, per its coding sessions page. Our Linear coding sessions explainer goes further.
Sinatra
Sinatra doesn't bill a mention at all. A task started from Slack counts against the workspace's plan like a task from Linear or GitHub, and the plan is a flat price for the whole workspace. Free is 1 task at a time, plus 1 queued, with no daily limit and no card. Pro is $20 a month: 3 tasks at a time, plus 10 queued, any number of people. The model work runs on the Claude or ChatGPT subscription of the person who mentioned it, or on a workspace API key, and Sinatra never bills for model usage.
Side by side
A Slack bot puts the start button in front of everyone in the channel who's allowed to use it, so it's worth knowing which way the bill goes when the channel gets busy. Claude Tag, ChatGPT and Linear draw on a balance the organization funds, and a busy channel drains it. Cursor and Copilot put tokens on their own invoice, and so does Devin once its allowance runs out. One long session there can cost more than you expected. Sinatra and the older Claude app put model use on a subscription someone already pays for, so the person who typed the mention uses up their own quota. On Sinatra a workspace can also use a shared API key instead, and then your provider bills per token.
Slack Code and code channels
Slack's developer docs describe Slack Code as dedicated channels for working with agents, where "one channel is created per task, and they're archived when the session finishes." Each one has view tabs the agent manages, for diffs, HTML, Block Kit or canvases. Slack's help article says it's available on all Slack plans with a supported agent installed, is "rolling out gradually," and lists Claude, Devin, Factory and GitHub Copilot among the supported agents. GitHub documents the most here: anyone who joins the code channel can "add context, redirect the approach, or stop the session."
Sinatra doesn't support code channels. A Sinatra task lives in the thread where it was mentioned.
Where the other Slack agents win
Sinatra has no DMs, no stop control in Slack, no Teams app, no code channel support and no Slack Marketplace listing. If any of those matters to your team, one of the others is the better pick.
Claude Tag, ChatGPT, Copilot and Factory document direct messages, and Linear takes group DMs, so developers who want to ask for something privately can.
Claude Tag has a Stop button on its working indicator, Cursor's context menu has a delete that stops and archives the agent, Devin takes EXIT and sleep, and Copilot can be stopped from its code channel. A Sinatra task is stopped from the dashboard, sinatra stop, or the MCP or HTTP API.
ChatGPT, Cursor, Devin, Copilot and Factory document a Microsoft Teams app, Linear's agent works there too, and Anthropic has a waitlist for Claude Tag in Teams. Claude, Devin, Factory and Copilot are in Slack's list of code channel agents, and Copilot lets anyone in the channel redirect the session.
Claude Tag doesn't need a seat per person: "Slack users don't each need a Claude seat to work with Claude in channels." For a company already on Claude Team or Enterprise, it puts Claude in channels without anyone installing anything.
Linear starts a coding session from a Slack thread with nothing new to install if you're already on a paid Linear plan. Linear is also one of the few vendors here that publishes how a session is priced.
Claude, Devin, Cursor, Factory and OpenAI's ChatGPT app are listed in the Slack Marketplace. You connect Sinatra's Slack app from workspace Settings in Sinatra instead.
Where Sinatra fits
In Sinatra, each Slack thread is one task. You mention @Sinatra in a channel or inside an existing thread, it reacts with 👀, keeps one progress checklist in the thread, and posts its answer, question or plan as a reply. When the request needs code, it opens a pull request and links it back. A reply that mentions it always continues the task, and an unmentioned reply does when Sinatra judges it's meant for it. A message sent while the agent is running reaches it, the same as on Linear and GitHub. After the PR merges, the same thread still takes follow-ups, and the task can open another PR.
Each member runs Slack tasks on their own Claude or ChatGPT subscription, or the workspace uses a shared API key. Sinatra never bills for model usage. The Slack app is on every plan including Free, and each plan is one price for the whole workspace however many people mention it. For a small team the cheapest honest path is the subscriptions you already pay for, plus $20 a month for the whole workspace.
It's also the same task model as the rest of Sinatra, so the label on a GitHub issue, an assignment in Linear and a mention in Slack all end up as tasks on one dashboard. The Slack page has the overview and the Slack docs have the setup.
How to choose a Slack coding agent
Start with what your company already pays for. On Claude Team or Enterprise, Claude Tag is built for answering questions in channels as much as for writing code, and the spend limit is the control to set first. On ChatGPT Business or Enterprise, @ChatGPT is already there once an admin turns it on, with Codex cloud behind it for repository work. A team that lives in Cursor gets a Slack app that shares their Cursor setup and bills the same way the editor's cloud agents do.
Copilot Business and Enterprise have the most complete version of code channels with stop and redirect. GitHub still calls its new experience a public preview.
Sinatra suits a team that wants the agent's model use on subscriptions it already holds, one flat workspace price, and the same task model across Linear, GitHub and Slack. Its gaps are in the table above. If you start work from claude.ai or ChatGPT rather than Slack, the chat post covers that path, and the background agents roundup and the alternatives hub cover the rest.
Common questions
Is the Codex Slack bot still @Codex?
No. OpenAI now says to start new requests with @ChatGPT once an admin enables the ChatGPT app, and that repository work goes to a separate Codex cloud task when cloud delegation is on.
Which Slack coding agents take direct messages?
Claude Tag, ChatGPT, Copilot and Factory document DMs. Linear takes group DMs. The older Claude Code in Slack app and Sinatra work in channels only.
Does a Slack mention cost money on its own?
On most of these, yes, through whatever meter the agent uses: an organization balance, AI credits, tokens at API prices or a plan allowance. On Sinatra, a mention that starts a new task counts against the plan's tasks-at-a-time limit, a mention in a thread that already has a task is a follow-up on it, and the model use goes to your own subscription or key.
Can I stop a coding agent from Slack?
Claude Tag, Cursor, Devin and Copilot (in code channels) document a way to. ChatGPT shows a Cancel control when the deployment allows it. Sinatra doesn't; stop its tasks from the dashboard, the CLI or the API.
Which agents work in Slack code channels?
Slack lists Claude, Devin, Factory and GitHub Copilot among others. Sinatra doesn't support them.
If you want to see what a Sinatra thread looks like, start free, have an owner or admin connect Slack from Settings, and mention it under the next bug someone posts. The Claude Code, Codex, Cursor, Devin, Copilot, Factory and Linear comparison pages have the rest of each product, side by side with ours.