An AI agent platform with Slack integration is one that can read and write inside a Slack workspace, but the phrase hides a fork. Either the agent lives in your channels as a participant that answers mentions and direct messages, or it runs somewhere else and posts into a channel when a job finishes. The first changes how a team works. The second saves somebody from copying a report.
Nearly every platform in this list will answer yes to “do you integrate with Slack”. The useful question is a different one: when a colleague types a question in a channel at 4pm, does the agent read it and reply, or does the message sit there until someone runs a workflow by hand?
The distinction decides three practical things. It decides the price, because a participant is metered by conversation and a printer is metered by run. It decides the security review, because a participant needs scopes to read your messages and a printer usually needs only permission to post. And it decides whether your team adopts the thing at all. A bot in a channel is used. A workflow dashboard is visited when somebody remembers.
Shogo connects to Slack through OAuth and also ships its own agent-native team chat, so it sits on the participant side of that fork. It appears at number 10 below, after the nine competitors reviewed on the same four points.
The two things people mean by Slack integration
A participant. The agent has a bot identity in the workspace. You mention it in a channel, or open a direct message, and it answers with the context of that conversation. It can hold a thread, offer buttons, and remember what was asked two messages ago. Under the hood this is a Slack app with event subscriptions, bot scopes and a request URL that Slack calls. Gumloop, Lindy, Relevance AI and Zapier Agents all work this way, and so does a Shogo agent once Slack is connected.
A printer. The agent or workflow runs on its own schedule and calls Slack’s API to post a message. Nothing in Slack can start it, and nobody can ask it a follow-up in the channel. Make, n8n, Activepieces, Power Automate and Dify model Slack this way: the Slack app is a set of modules or tools that a flow calls, and the trigger is either a poll or a webhook that fires a workflow rather than a conversation.
Both are legitimate. A daily revenue summary belongs in the second category and needs no bot identity. An agent that triages support questions belongs in the first, and a printer cannot fake it: without event subscriptions there is nothing to answer.
What changed in 2026: Slack now has its own agent layer
On 20 August 2026 Slack launched Add to Slack, a one-click deployment path built with ten partner platforms: Hyperagent, LangChain, Lovable, n8n, NanoClaw, OpenAI, Runlayer, Skydive, Superhuman and Vercel. Slack’s own announcement says a deployed agent “automatically inherits Slack’s enterprise permissions and data boundaries”, and that agents built on partner platforms move into a workspace without custom integration code.
That matters when you compare vendors, because it separates three tiers. A platform on that partner list has a sanctioned deployment path. A platform with its own Slack app or custom-app route has a working integration that your IT team has to review. A platform that only posts messages is not in this conversation at all, whatever its marketing page says.
Slack’s own framing is also worth reading closely: the announcement describes business teams getting automation in the flow of work while IT keeps “full visibility, permission controls, and data sovereignty”. Whether that holds depends on which scopes an app requests and where it sends what it reads, which is a question for the vendor rather than for Slack.
The limit nobody quotes
Slack is not an unlimited message bus, and an agent is a message generator. Slack’s rate limit documentation states that chat.postMessage “generally allows posting one message per second per channel, while also maintaining a workspace-wide limit”, and that Tier 3 methods, which most messaging calls belong to, allow 50 or more requests a minute.
An agent that posts one update per task breaks none of those rules until the volume arrives. Fifty tasks an hour into one channel is a message a minute and fine. A digest agent fanning out to forty channels, or a monitoring agent reacting to every alert in a busy workspace, meets the ceiling sooner than people expect.
Vendors inherit that ceiling, and some publish numbers of their own. Microsoft’s Slack connector documentation lists a throttling limit of 100 API calls per connection per 60 seconds, and states that the trigger polls every 120 seconds. Make’s documentation notes that Slack tightened the rate limits on the conversations.history and conversations.replies methods, and that its Watch modules now expose a pagination toggle to reduce API calls when those errors appear.
The practical consequence: when an agent feels slow in Slack, the bottleneck is usually the connector’s polling interval or Slack’s own limit, not the model. Ask for both numbers before you buy.
Five checks before you connect anything
1. Participant or printer. Read the vendor’s Slack page and look for the words mention, direct message, event subscription or slash command. If all the examples end with “post a message to a channel”, you are buying a printer.
2. Whose identity the bot wears. Some platforms put agents behind a shared vendor bot, so every customer’s agent appears as the same name. Others let you create your own Slack app with your own name and avatar. If multiple agents share a workspace, the shared-bot model gets confusing fast.
3. What leaves the workspace. A participant reads messages, which means message content reaches the vendor’s model. That is usually acceptable and always worth writing down: which scopes, what retention, whether the data trains anything, and whether a self-hosted deployment is available for the channels that carry customer data.
4. What 200 messages a day costs. Credit-metered platforms price a chatty agent like a chatty API client. Seat or execution-priced platforms do not. Our comparison of AI agent pricing models covers the mechanics.
5. Who administers the Slack app. Some organisations restrict app installation to admins or to apps from Slack’s marketplace. Make documents that it is not itself listed in the Slack Marketplace, which is a five-minute problem in a permissive workspace and a procurement conversation in a locked-down one. Check the policy first.
The 10 best AI agent platforms with Slack integration in 2026
Each platform is assessed on the same four points: what it is, where it is strong, where it falls short, and who it suits. Prices come from each vendor’s own pricing page in October 2026, and where a vendor publishes none in a readable form, this article says so.
1. Zapier
What it is: Zapier is a no-code automation platform, and Slack is one of its deepest integrations: triggers and actions across messages, files, reactions, channels and users, plus a documented Agents connection.
Strengths: The trigger list is the widest here. New message in any public channel, in a specific channel, in a private channel or multi-person direct message, a keyword or mention search, a new reaction, a new file, a new member, a new channel, a new custom emoji, and a message pushed from Slack. Actions cover posting, thread replies, uploads and channel administration. Zapier also publishes a Slack connection for its Agents product, so an agent can answer inside Slack rather than only at the end of a Zap. It reaches 9,000 other apps. Pricing is $0 on Free, Professional from $19.99 a month and Team from $69 a month.
Weaknesses: Work is billed in tasks, so an agent replying in a busy channel spends metered allowance on every message and every trigger poll. Deeper logic pushes you into paths, which are metered too. The shared Slack app model means your agent is a Zapier-shaped bot rather than your own named app.
Best for: teams already standardised on Zapier that want Slack as one channel among many.
2. n8n
What it is: n8n is a source-available workflow automation platform with a Slack action node and a separate Slack Trigger node.
Strengths: The Slack Trigger node verifies that incoming requests “are from Slack and include a trusted signature”, and n8n recommends keeping that on. That is the difference between a workflow that reacts to Slack and one that can be made to react to anyone who finds its webhook URL. The action node covers channel, message, file, reaction, user and user-group operations, and when an operation is missing you drop to an HTTP Request node instead of waiting for the vendor to ship it. Self-hosting keeps message content inside your own infrastructure, which is usually the reason a regulated team picks n8n. Slack named n8n among the ten launch partners for Add to Slack.
Weaknesses: The pricing page did not return a readable set of figures to this review, so confirm the tier before you model the cost. Self-hosting moves patching, uptime and secrets onto your team. The trigger node needs a public endpoint, so a Slack-triggered workflow cannot sit entirely behind a closed firewall unless you front it with something that is reachable.
Best for: teams with engineering capacity that want the flow and its data on their own machines.
3. Make
What it is: Make is a visual automation platform with a Slack app of modules covering messages, files, channels, reactions, users, reminders and statuses.
Strengths: The instant Watch New Events module uses a webhook that Make attaches itself, so there is no manual webhook setup in Slack. The module documentation is specific about connection types and about which API method each operation calls, which makes the integration predictable to debug. Actions and triggers are separated cleanly from the rest of the scenario, so a Slack step reads like any other step in a visual flow. Make lists Core from $9 a month and Pro from $16 a month.
Weaknesses: Make documents two sharp edges itself. Watch modules only work with user connections rather than bot connections, which means the automation acts as a person. And Make is not in the Slack Marketplace because Slack has not approved it, so a workspace that restricts installs will block the connection until an admin intervenes. Its pricing page returns an error to automated requests, so confirm the figure at checkout rather than trusting a list price quoted anywhere, including here.
Best for: operations teams that want a visual scenario and modules they can reason about line by line.
4. Power Automate
What it is: Power Automate is Microsoft’s automation service, and Slack is a Standard-class connector available in Power Automate, Logic Apps, Power Apps and Copilot Studio.
Strengths: Microsoft publishes the operational numbers buyers actually ask for. The connector allows 100 API calls per connection per 60 seconds, and the trigger polls every 120 seconds. Being Standard class means it does not require a premium connector licence, which is unusual for a third-party service inside the Power Platform. Flows inherit the tenant’s identity, conditional access and data loss prevention policies, and that governance story is often the deciding factor in an organisation that already runs Microsoft 365.
Weaknesses: A 120-second poll floor makes this a reporting pipe, not a conversation: a reply to a Slack message can wait two minutes, and nothing in Slack can address the flow directly. There is no bot identity to mention. The connector documentation also marks older posting actions as deprecated in favour of Post message (V2), so flows built from old templates need revisiting. The product page did not return a readable list price to this review.
Best for: Microsoft-standardised organisations that need Slack inside a governed flow rather than as the interface.
5. Activepieces
What it is: Activepieces is an MIT-licensed, open-source automation builder whose Slack piece ships 29 actions and 14 triggers.
Strengths: The trigger set is unusually complete for an open-source tool: new public message anywhere, new message in a chosen channel, new direct message, mention in a channel, mention in a direct message, reaction added, reaction removed, new channel, new command in a channel or direct message, new user, new saved message, custom emoji added and modal interaction. Two actions handle approvals directly: request approval from a user and request approval in a channel, both of which send approve or reject buttons and wait for the answer. Pricing is $0 on Free, $20 a month on Plus and $200 a month on Team, with $0.007 per credit beyond that, and the piece is open source so it can run on the vendor’s cloud or on your own.
Weaknesses: The approvals are flow steps, not a conversation, so a person clicking a button ends that branch rather than continuing a thread with the agent. Piece quality across the 770-plus apps varies, and a self-hosted install means you own upgrades. The interactive pieces of a Slack app still need the Slack app configured, which is developer work however friendly the builder is.
Best for: teams that want the Slack plumbing open source, cheap and inspectable.
6. Gumloop
What it is: Gumloop is an agent builder with two documented Slack routes: a standard @Gumloop integration and a custom Slack app you own.
Strengths: The documentation separates the two clearly, and it is the most honest explanation of this trade-off in the list. The standard integration is fastest, works per channel, supports Block Kit messages and buttons and slash commands. The custom app gives the agent your own name and avatar, direct messages, several agents in one workspace under different identities, white-label deployments, and a route for IT policies that require apps to be owned internally. Gumloop also documents the scopes and event subscriptions involved, so a security reviewer has something concrete to read. Pro is $37 a month after a 14-day trial, with Enterprise priced on request.
Weaknesses: The standard integration is scoped per channel, so an agent in ten channels is ten connections. Set-up for the custom app is a documented 10 to 15 minutes and needs Slack admin rights, and the custom app does not support slash commands, so the two routes do not have identical features. Credit-based pricing means a conversational agent’s cost tracks message volume.
Best for: teams that want an agent answering in Slack under their own bot identity.
7. Lindy
What it is: Lindy is an agent platform whose agents can work inside a Slack channel, which is the participant model rather than a workflow that posts.
Strengths: You add the agent to a channel and ask it to take work off your plate in plain language, which is the lowest-friction version of this integration: no trigger to configure and no flow to draw. The action set behind it is broader than the chat surface suggests, covering channel creation and archiving, Block Kit formatted messages to channels, users or groups, scheduled messages, and direct messages. For a team that already thinks in Slack threads, the mental model fits on day one.
Weaknesses: Credits are consumed by conversation, and top-ups run at $10 per 1,000 credits, so a chatty agent’s bill tracks messages rather than seats. The pricing page lists plans from $29.99 a month for 4,000 credits, shown against a $39 list price, with $50 of free credits to start, and promotional pricing moves, so confirm the number at signup. Slack is one surface among several rather than the product, which means per-channel controls are thinner than a dedicated Slack app gives you.
Best for: small teams that want a colleague in the channel more than they want a workflow.
8. Relevance AI
What it is: Relevance AI is an agent platform that connects Slack as a tool set for its own bot user, with 26 pre-built Slack tool steps.
Strengths: The inventory is published by name: 24 actions and 2 triggers. Actions include threaded replies, a large-message tool for anything over 3,000 characters, channel topic and description updates, user-group membership, and listing replies or channel members. The two triggers are a new message in a channel and a new direct message, so an agent can start work from Slack rather than only finish there. Connection is one bot token pasted per workspace, after which every agent in that workspace can use it.
Weaknesses: The bot token is the security model, so custody of that token is yours to manage, and rotating it is a manual job. Pricing is enterprise only, listed as Custom, with no published entry tier, which makes it hard to trial without a sales conversation. Slack reads happen through tool steps an agent chooses, not through a scheduled sync, so anything needing guaranteed ingestion has to be driven by the agent.
Best for: teams already running Relevance workforces that want Slack as both an input and an output.
9. Dify
What it is: Dify is an LLMOps platform where Slack support comes from a marketplace plugin rather than a native channel.
Strengths: The Slack plugin, verified by Dify and authored by the platform’s own team, exposes about 20 tools: send, update, delete and schedule messages, fetch a permalink, post through an incoming webhook, list and create channels, read channel history and thread replies, invite users, add and remove reactions, look up users, and upload files. It handles both the bot token path and the simpler webhook path, so a one-way notification needs no scopes at all. Self-hosting is available, and pricing is $590 per workspace a year for Professional with 5,000 message credits a month, or $1,590 for Team.
Weaknesses: Slack here is a tool, not a surface. You build the reply loop yourself, and nothing answers a mention unless you also wire up an endpoint. Message credits are consumed by the work behind each call, so a busy Slack app draws on the same pool as every other app. A Slack app and bot token are still required for anything beyond webhook posting.
Best for: teams already building on Dify that want Slack as one of several outputs.
10. Shogo
What it is: Shogo is an open-source AI agent harness combining team chat, meeting memory, software building and every frontier model, with 1,018+ integrations including Slack.
Strengths: Slack connects by OAuth, with the agent walking you through it: ask it to connect, click the link, authorise. No token is pasted and no manifest is edited. Shogo also ships its own agent-native team chat, so Slack can be a bridge to a legacy workspace rather than the only place work happens. Its templates cover the common cases: support triage with escalation, operations monitoring that tracks Slack mentions, and a standup digest built from GitHub and Slack activity. Paid plans are unlimited inside rolling fair-use windows, so an agent posting 200 updates a day costs the same as one posting two. Deployment can be Shogo’s cloud or your own VPC.
Weaknesses: Slack is one of 1,018 integrations rather than a Slack product, so there is no per-channel agent console and no slash-command builder of the kind Gumloop’s custom app documents. A team that wants a different bot identity per channel will find Gumloop’s route more explicit. Free and Basic are single-user, so a team starts at Pro at $20 a seat.
Best for: teams that want Slack as one surface among several, at a flat seat price.
How to choose
If people should ask the agent questions in a channel, pick a participant: Lindy, Gumloop, Relevance AI, Zapier Agents or Shogo. Then check identity and scopes, because that is where the security review lands.
If a report just needs to appear, pick a printer and stop paying for conversation features you will not use. Make and Power Automate are the clearest to reason about, n8n and Activepieces are the ones you can self-host.
If your workspace restricts app installs, read the vendor’s own documentation about marketplace approval before you promise anyone a demo.
If the volume is high, model the two meters together: the platform’s credits or tasks, and Slack’s one-message-per-second-per-channel limit. Our roundup of AI workflow automation tools for operations teams covers the workflow side of that bill.
Where Shogo fits, and where it doesn’t
Shogo is the right pick when Slack is one of several surfaces and the agent’s job crosses systems. The OAuth connect is the shortest setup in this list, flat-rate pricing means a talkative agent does not run up a bill, and the shipped Slack-adjacent templates cover the operations cases most teams start with. The AI agent guardrails patterns for approvals and audit logs apply directly to what an agent does after it reads a message.
It is the wrong pick if Slack is the product you are buying. There is no screen that maps agents to channels, no slash-command builder, and no per-channel identity, so a support organisation that wants four different bots in four channels should look at Gumloop or a custom Slack app. Teams that need every message to stay inside their own Slack instance for retention reasons should also ask harder questions of any hosted agent, Shogo included, than a marketing page can answer.
Frequently asked questions
What does Slack integration mean for an AI agent?
It means the agent can read or write inside a Slack workspace through Slack’s APIs. The important variable is direction: a participant agent receives mentions and direct messages and replies in a thread, while a workflow integration only posts messages at the end of a run. Two products can both claim Slack support and behave nothing alike.
Can an AI agent reply to a Slack message, or only post one?
Replying requires event subscriptions, bot scopes and a reachable endpoint, which is the participant model used by Lindy, Gumloop, Relevance AI and Zapier Agents. Posting requires only a webhook or chat.postMessage, which is what Make, n8n, Activepieces, Power Automate and Dify do out of the box.
What are Slack’s rate limits for an AI agent?
Slack’s documentation states that chat.postMessage generally allows one message per second per channel while also enforcing a workspace-wide limit, and that Tier 3 methods allow 50 or more requests a minute. Connector-level limits sit on top: Microsoft’s Slack connector allows 100 calls per connection per 60 seconds and polls triggers every 120 seconds.
Do I need a custom Slack app for my agent?
Not for a single agent in a few channels. You need one when several agents must appear under different names in the same workspace, when IT policy requires internally owned apps, or when you want direct messages answered by a specific agent. Gumloop documents both routes, and its own comparison is the clearest public write-up of the trade-off.
Is it safe to give an AI agent access to Slack?
It is as safe as the scopes you grant and the token you store. A participant agent reading channels needs message history access, so message content reaches whichever model the platform runs. Ask which scopes are requested, whether the app can be scoped to specific channels, where the data is processed, and whether a self-hosted or VPC deployment is available.
Which platform is best for an agent that lives inside Slack?
Lindy is the simplest to start with because you add it to a channel and talk to it. Gumloop is the most explicit about identity and enterprise requirements. Shogo is the best value if the same agent also has to touch Jira, a database or a build pipeline, because flat-rate pricing keeps message volume out of the bill.
Can an agent read every channel in the workspace?
Only if the app is granted the scopes to do it and the bot is invited to private channels. Public channel history needs channels:history or equivalent, private channels need separate scopes and an invite, and direct messages need im:history. Most vendors request the minimum for their documented features, and the app’s OAuth screen shows exactly what is being asked for before you approve it.
Sources
- Zapier, Slack integrations page and pricing page
- n8n, Slack node and Slack Trigger node documentation
- Make, Slack app documentation and pricing page
- Microsoft, Slack connector reference for Power Automate, Logic Apps, Power Apps and Copilot Studio
- Activepieces, Slack integration page and pricing page
- Gumloop, custom Slack app documentation and pricing page
- Lindy, Slack integration page and pricing page
- Relevance AI, Slack integration page and pricing page
- Dify, Slack plugin on the Dify Marketplace and pricing page
- Slack, Add to Slack announcement (20 August 2026) and rate limit documentation
- Shogo, integrations directory, why-shogo, pricing and llms.txt
