Merlin — Claude Code and Codex Agents in Slack, Linear, and GitHub

What is Merlin?

Merlin is a collaboration tool that turns personal AI coding agents into shared team resources. Its tagline is "Your agents, now part of your team": instead of each engineer driving Claude Code or Codex alone in a terminal, Merlin connects those agents to the channels where team communication already happens — Slack, Linear, and GitHub. It is built by Lovin, LLC and aimed at "small AI-native teams" across engineering, product, marketing, support, people, finance, and data functions. Importantly, Merlin does not sell model access: it orchestrates your existing Claude Code and Codex subscriptions running on your own computers.

How Merlin Works

Setup is described in three steps. First, connect one machine via the Merlin desktop app or CLI and sign in with "claude login" and "codex login" — the agents keep running on your hardware, using your accounts. Second, add Merlin to Slack channels and choose its mode: mention-only, proactive, or thread follow-ups. Third, ask for work in a thread and results come back to the same place. Beyond ad-hoc requests, Merlin supports recurring Schedules, a shared workspace skill catalog, access control, and workspace/channel Memory so agents retain context. The site stresses "your machines, your accounts" — no API keys and no separate model bill, with model usage billed through your existing Claude and Codex subscriptions.

Main Features

• Slack, Linear, and GitHub integration — agents accept work from the channels teams already use, including PR feedback and task updates.
• Mention-only and proactive modes — respond when invoked, or take initiative when it matters.
• Shared skills catalog — workspace-wide skills every connected agent can use.
• Schedules — recurring tasks run automatically on a schedule.
• Access control and memory — permission scoping plus workspace and channel memory for context retention.
• Self-hosted execution — agents run on your connected machines under your existing Claude/Codex accounts.

Pricing

Merlin uses a freemium, per-workspace model. Free costs $0 for up to 5 active users with 1,000 requests and schedule runs per month and includes the Slack, Linear, and GitHub integrations. Team costs $59 per workspace per month for up to 50 active users with unlimited requests, priority support, and a 14-day free trial that requires no card. Scale costs $99 per workspace per month for up to 200 users. Larger organizations are asked to contact support. One plan covers the whole workspace with no per-seat fees, and a "request" counts every task Merlin runs — Slack mentions and DMs, Linear and GitHub requests, follow-ups, and scheduled runs. Active users include unregistered Slack/Linear/GitHub requesters; bots do not count. Free workspaces that exceed limits have new requests refused until the period resets or the workspace upgrades.

Common Use Cases

• Engineering — turning PR feedback and bug reports into code changes without leaving GitHub or Slack.
• Product — drafting PRDs and specs by mentioning the agent in a channel.
• Support and operations — scheduled runs for recurring checks and reports.
• Cross-functional teams — giving marketing, finance, and data teammates agent access through mentions rather than terminals.

Limitations

Merlin supports only two agents — Claude Code and Codex — so teams standardized on other tools are out of scope. Agents execute only on machines you connect, which means availability tracks your hardware staying online rather than a cloud service's SLA. Model usage stays on your own subscriptions, so heavy agent use consumes your Claude or Codex quota, and the Free tier's 1,000-request cap can bind quickly for active teams. Merlin only sees Slack channels it is invited to. The site references FAQ answers on security, data access, and cost that are not visible on the page, and no country or launch date is published.

Summary

Merlin addresses a real gap for small teams already paying for Claude Code or Codex: those agents sit in individual terminals while team requests arrive in Slack, Linear, and GitHub. By routing requests to agents on your own machines — with workspace pricing that stays free up to five people and no per-seat fees — it is a low-friction way to share agent capacity. The model is correspondingly limited: two supported CLIs, self-hosted availability, your own model quotas, and an early-stage product whose security documentation is not yet on the site. It fits AI-native teams of up to a few dozen people best.