Kadmo in Jira, Slack and Teams, light mode and reporting

The first release opened app.kadmo.ai on 7 October. This one is about where you meet Kadmo: in the tracker and the chat you already use, in the light or dark your system runs, and in what comes back to you once an agent has worked. Everything here reaches every account at once; there is nothing to install or upgrade on your side.

Ask Kadmo in Jira, Slack and Teams

Write @kadmo and what you need in a comment on a ticket, or add the rule’s label, and Kadmo decides which of the playbooks and workflows your workspace allows the request is asking for, starts it, and answers on the ticket. When the work ends, the ticket gets a ✓ or an ✗ and what came of it. If Kadmo cannot tell what you mean, it asks, and your next comment is the answer. How this compares with one coding agent per developer is on the Codex alternative page.

A message in a registered Slack channel or a Teams chat goes through the same decision, against the channel’s or the workspace’s list, and the answer follows in the Slack thread or the Teams chat. In Teams you confirm before anything starts.

Each workspace lists what a request may start and who may ask. Under Team › People, a person’s Jira and Slack users are linked to them, and their own list can be narrower than the workspace’s. The decision is two small calls on your account’s provider; without a provider, Kadmo answers with its menu instead.

Kadmo LLM decides in the docs

The app in light and dark

Every page of the app now stands on one design system, in Light and in Dark: the fleet, jobs and tasks, runs, workspaces, playbooks, chat, goals, settings and integrations share the same type, components and frame. From this release the app follows your system’s setting, and under Settings › Appearance each member picks System, Light or Dark for their own browser. The choice applies at once.

Appearance in the docs

Reporting: what comes back, and where

Task setup › Reporting is one page for what comes back from the work. Agent reports set the shape of the report an agent writes on a run’s ticket and pull request, its details and a preview, for the account, and a playbook step can set its own. Kadmo’s notices set what Kadmo itself writes at Started, Needs you and Done, on the ticket, in Slack, in Teams and by e-mail, with the presets Standard, Only when needed and Silent, and a workflow or a playbook can override them. Each person turns their own Slack DMs, Teams DMs and e-mails on or off under Settings › My notifications. Nothing changes until an admin changes it.

A report also renders where people read it: a shared report link opens as a page, and the run and job pages show each step’s report.

Reporting in the docs

More models from one account

The GPT models serve from Kadmo’s own OpenAI project in the United States, on the Kadmo provider and in the picker with where they run. An Enterprise account can connect its own OpenAI or Anthropic key instead, and every chat, chat agent and agent job then runs on it, through the same gateway, at no Kadmo price. The models and their prices are on the pricing page.

Models and credit in the docs

The Kadmo Agent Runtime and its MCP server

The Kadmo Agent Runtime is the part of Kadmo that runs on a computer. Every Kadmo agent runs it, and now you can too: it is one npm package, @kadmo/cli, holding the kadmo command, the agent service with its dashboard, and the MCP server.

npm install -g @kadmo/cli
claude mcp add kadmo-agent -- kadmo mcp

Register kadmo mcp as kadmo-agent with Claude Code, Claude Desktop, Codex, Cursor, VS Code, Windsurf or Gemini CLI. The client starts it itself, over stdio, with no token: it can then run your workflows, read your skill packs and their role guides as resources, upload artifacts, and drive your Chrome through the Kadmo extension. Beside it, kadmo remains your account’s MCP server at app.kadmo.ai/mcp, so one client holds both: your account and the runtime on your computer.

The package is compiled and ships no packs. They come from the Kadmo app, the public catalogue, a registry, a git repository or a folder, each installed once per domain with its layer, so an account’s pack extends the library’s and never fights it. kadmo serve starts the agent service on localhost:9876 with the dashboard: the installed packs, the workflows, the run logs and the settings.

The Kadmo runtime in the docs

The documentation at docs.kadmo.ai

Kadmo’s documentation has its home: the app page by page, agents, workflows and playbooks, skill packs, integrations, the MCP servers, models and credit, and the runtime.

docs.kadmo.ai

Also in this release