Most teams already use AI agents at work, but the chat app they sit in usually wasn't built for them — an agent is a bot account bolted onto Slack, its output copy-pasted in by whoever ran it, its actual reasoning and permissions invisible to everyone else. A newer wave of team chat tools flips that assumption: agents get a real seat at the table, with their own identity, permissions and a visible history of what they did, sitting in the same channel as the humans instead of off in a private terminal.
The short version
The short version: want a free, self-hosted Slack replacement you fully control? Chatto is a single-binary app you run on your own server. Want the same idea but decentralized, open-source and backed by Jack Dorsey's Block? Buzz gives AI agents like Claude Code and Codex a real cryptographic identity in the chat itself. For a team that wants pre-built agent roles (engineer, PM, marketer) with full audit logging, Vokal goes furthest. SquidHub is the lighter option — real-time shared rooms where humans and AI "squids" brainstorm together using your own API keys. If your team already runs AI coding agents individually and wants to see what they're doing, CoPatch puts that in a shared workspace. And if your team is a construction or field crew rather than a software team, Convoe turns group-chat conversations into tracked tasks automatically.
Chatto — self-hosted Slack, one binary, your own server
Teams that want Slack-like chat but don't want every message and file routed through a third party's servers usually end up choosing between clunky self-hosted options or just accepting the trade-off. Chatto is built squarely for that gap: channels, voice and video calls, and screen sharing, running as a single-binary TypeScript app you deploy yourself, in the EU or wherever you choose.
Pricing: free, self-hosted, open-source (AGPL-3.0 core, Apache-2.0 for UI and integrations). Cloud hosting is announced as coming but not yet available.
Forces: genuinely simple deployment (no separate database needed for a basic setup), voice/video and screen sharing included out of the box, and active development — about 2,300 GitHub stars and 1,180+ commits since open-sourcing in July 2026.
Limites: young project with fewer integrations than established alternatives, AGPL licensing implications worth checking for your use case, and external contributions are currently closed while the core team stabilizes.
Pick Chatto if: you want a Slack-shaped chat app you fully own and host, with no monthly per-seat bill.
Buzz — open-source, decentralized, agents as real teammates
Most team chat tools treat AI agents as an afterthought. Buzz, launched by Block CEO Jack Dorsey in July 2026 explicitly to cut teams' dependency on Slack and GitHub, starts from the opposite assumption: agents get their own identity, their own permissions and a visible audit trail, in the same conversation as human teammates rather than off in a sidebar.
Pricing: free and fully open-source (Apache 2.0).
Forces: built on the decentralized Nostr protocol so there's no central server to trust; already supports agents like Goose, Codex and Claude Code with cryptographic identity; combines channels, DMs, voice, file sharing and Git repositories with automated workflows triggered by messages or events.
Limites: no mobile app at launch (desktop only — macOS, Windows, Linux), very new, and the decentralized model is a different adoption curve than a typical SaaS signup.
Pick Buzz if: you want a genuinely free, open-source Slack alternative where AI coding agents are first-class participants from day one.
Vokal — pre-built agent roles with full audit logging
AI coding agents make one developer faster individually, but if an agent's work only lives in that person's private terminal session, nobody else on the team can review it, question it, or build on what it learned. Vokal gives that work a shared home — channels, tasks and a knowledge base — with every agent action logged and visible rather than hidden.
Pricing: not published; quote-based.
Forces: 20+ pre-built agent roles (software engineer, PM, marketer, support, ops), permission scoping per user and per agent, and integration with 1,000+ apps via Composio plus direct integrations for Slack, Linear, Jira, GitHub, Notion and Google Docs.
Limites: no public pricing to evaluate upfront, the company's funding and track record aren't disclosed, and it's really built for a team with enough scale to need role-based agent structure.
Pick Vokal if: your team runs multiple AI agents across different functions and needs a real audit trail of what each one actually did.
SquidHub — real-time rooms, bring your own model
The default AI workflow today is everyone running their own private ChatGPT or Claude conversation, then copy-pasting the useful bits into Slack — losing the context behind the answer in the process. SquidHub replaces that with shared real-time rooms where teammates and their AI agents ("squids") brainstorm, plan and write together in the open.
Pricing: not publicly listed — self-funded with no venture capital, likely freemium or usage-based on request.
Forces: works with hosted Claude access or your own API key for Claude, GPT, Grok or Gemini; reusable instruction bundles attached to any squid; content encrypted at rest (AES-256-GCM); and explicit that Anthropic applies zero data retention on requests routed through it.
Limites: no public pricing tiers, a young product with limited history, and less funding firepower than venture-backed competitors.
Pick SquidHub if: you want lightweight, model-agnostic shared AI rooms without committing to one AI vendor.
CoPatch — shared visibility into your team's AI coding agents
As more developers bring AI coding agents into daily work, those agents typically end up siloed in each person's own terminal or IDE, invisible to everyone else. CoPatch puts people and agents like Claude and Codex in the same shared workspace, so the team can see and coordinate around what agents are doing instead of running disconnected individual sessions.
Pricing: unpublished — the product is currently in early access, waitlist-only.
Forces: targets a real, emerging problem (fragmented per-person agent workflows), combines human collaboration and AI agents in one space, and emphasizes keeping the user in control of what agents can do.
Limites: waitlist-only with nothing public to test yet, no pricing, and no GitHub repository to evaluate the actual implementation.
Pick CoPatch if: your team already runs Claude Code or Codex individually and you're willing to be an early adopter of a shared-visibility layer on top.
Convoe — for crews, not just software teams
On a construction site or in the field, coordination still mostly happens in a group chat, and it's easy to lose track of who actually handled what. Convoe keeps that familiar group-chat feel but adds an AI layer that reads the conversation and turns it into a tracked task with an owner, a due date, and a link back to the original message and photo.
Pricing: free core plan (unlimited crew, chat/tasks/calendar); Lite is $19/month, Pro is $39/month, plus $12/active AI user/month for AI features.
Forces: a genuinely free tier that covers real collaboration for unlimited users, multiple purpose-built AI specialists (summaries, actions, email, reporting, reminders), and source-linked records with timestamps and photos.
Limites: paid AI plans are per-user and can add up for large crews, and the construction/field positioning makes it a poor fit outside that niche.
Pick Convoe if: your team works on-site rather than at a desk and needs chat that automatically becomes a task log.
| Tool | Model | Starting price | Best for |
|---|---|---|---|
| Chatto | Self-hosted, single binary | Free | Owning your own Slack-like server |
| Buzz | Open-source, decentralized (Nostr) | Free | Agents as first-class, cryptographic participants |
| Vokal | Managed workspace | Quote-based | Structured agent roles + audit logging |
| SquidHub | Managed, bring-your-own-key | Not published | Model-agnostic real-time AI rooms |
| CoPatch | Managed, early access | Not published | Visibility into individually-run coding agents |
| Convoe | Managed, field-team focused | Free (AI from $12/user/mo) | Construction and field crews |
The common thread across all six is the same bet: that a team's AI agents shouldn't be invisible to everyone except whoever's running them. Whether you want to self-host that idea for free (Chatto, Buzz) or buy a managed version with more structure built in (Vokal, SquidHub, CoPatch), the right pick mostly comes down to how much control over your own infrastructure you actually need versus how much setup you're willing to do yourself.