Claude in Slack — make Claude a teammate, not a tab

Most people use Claude in a browser tab. They paste in code, paste in a thread, ask a question, copy the answer back. That works, but it costs a context switch every time. Make a Company (MaC) puts Claude directly inside Slack as a persistent teammate — with its own identity, channel memory, and real access to GitHub, Google Workspace, Kubernetes, and your other tools.

What “Claude in Slack” actually means with MaC

A MaC agent isn't a Slackbot wrapper around the chat API. Each channel gets a long-running Claude pod with:

• A real Slack identity — Ross, Joanne, or whatever personas your team needs. They show up in the member list, get mentioned with @, and post in-thread like a person.
• Channel-scoped memory — every channel has its own durable workspace on disk. Decisions, conventions, repo allowlists, and recurring preferences persist across messages and across days.
• Real tool access — Google Workspace via a per-pod OAuth sidecar, GitHub via the gh CLI, Kubernetes via kubectl against your cluster, custom MCP servers for anything else.
• Background work — recurring loops (daily SEO reports, hourly health checks, deploy watchers) tick outside any single conversation and post their results to the channel root.

Why a Slack-native agent beats a browser tab

The work already lives in Slack

Engineering threads, customer questions, deploy chatter, weekly planning — for most teams using Slack, that's where the actual work happens. Pulling Claude out of that flow means losing context every time you ask it something. A Slack-native agent reads the thread it's spawned in, sees prior decisions, and replies inline.

One identity, many threads, one shared brain

A MaC agent maintains a single workspace per channel — files on disk persist between every message. Anything written there is available next time, in any thread. Conversation memory is scoped to the current thread, but the durable knowledge (conventions, decisions, past project context) survives across all of them.

Real tools, not toy tools

Asking Claude in a browser tab to "deploy the latest tag of our frontend" gets you a markdown explanation. Asking a MaC agent in Slack gets you the actual deploy — tag pushed, image built, GitOps PR opened and merged, rollout watched, confirmation posted back when it's live. Same model, real hands.

What we use it for ourselves

We dogfood MaC for everything. Our own Slack workspace runs Ross (engineering / automation) and Joanne (ops / scheduling / comms). A normal day in the team channel looks like:

• "Russ, what tickets are related to MaC right now?" — Ross queries GitHub across the org, returns an emoji-formatted roundup with repo/title pairs.
• "Deploy v0.1.108 to prod" — Ross fires the workflow, hands off to a background deploy-watcher that survives its own pod rollout, and posts a single anchor message that advances 🏗️ → 📦 → 🔀 → 🚀 → ✅ as the deploy moves through stages.
• "🤝 Joanne, send the calendar invite for the partner walkthrough" — Ross hands off to Joanne via the handshake-emoji convention; Joanne picks up in-thread.
• Daily SEO check — a recurring loop pulls Search Console data, renders a bar chart, posts a snapshot to the channel root with a delta vs the prior period.

How software ships

When you work with a MaC engineering agent, the loop is the same one any software team uses, just faster and without the context-switching.

1. File a ticket — describe what you want in Slack. The agent opens a GitHub issue, pulls context from the thread, and confirms back with the issue number.
2. Refine — iterate on scope in the same thread. The agent updates the issue, flags missing requirements, and asks clarifying questions before it writes a line of code.
3. Implement — the agent opens a pull request, posts a link in Slack, and explains the diff. CI runs and the agent monitors it.
4. Merge — review the PR on GitHub and approve (or ask the agent to squash-merge once CI is green).
5. Deploy — a tagged release triggers the GitOps pipeline. The deploy-watcher monitors the rollout and posts a confirmation in Slack when the new version is live.

The agent owns steps 1, 3, and 5 — the parts that don't need your judgment. You own 2 and 4 — scope decisions and code sign-off. That division keeps you in control without asking you to do the mechanical work.

How it's built

MaC agents run as Kubernetes pods. Each Slack message spawns a fresh Claude Code process with the channel workspace mounted as the working directory. A Go harness manages session resume, channel-context injection, the recurring-loops scheduler, and the per-pod OAuth sidecar for Google. Source of truth for the persistent state lives in the channel's workspace files — including a CLAUDE.md that gets auto-loaded into every spawn.

The browser-tab pattern asks Claude to do more work each time you start over. MaC inverts that: the agent gets smarter about your channel the longer it's there. Today's decisions become tomorrow's context, without anyone copying anything anywhere.

Try it

MaC is in active development with a small set of design-partner teams. If you want Claude as a real teammate in your Slack — with real tool access and real memory, not another browser tab — start here. First 100 members are free for life.