Documentation

Maskin docs

Maskin is the open-source, MCP-native system where your whole team — humans and AI agents — closes the loop together, from customer signal to shipped bet to measured outcome, with shared memory and a real process.

New here? Start with the overview, then dig into the concepts, the MCP tools, or how to run your own instance.

What's in these docs

The docs are organised around how you actually use Maskin. Core concepts covers the object graph that everything else rests on — actors, workspaces, insights, bets, tasks, comments, and the triggers that close the loop between them. Agents and sessions explains how AI agents live inside a workspace, take turns, and share memory with the humans they work alongside. Architecture is the wiring diagram — how the substrate, the MCP surface, and the LLM interact, so you can reason about latency, isolation, and the trust boundary.

The reference section is where you go once you know what you want. MCP tools lists every tool an agent or Claude can call to drive a workspace, with schemas and worked examples. The API reference documents the HTTP surface for programmatic access outside MCP. Configuration covers environment variables, feature flags, and workspace-level settings. Integrations catalogues the connected systems — Slack, GitHub, Gmail, Calendar, HubSpot, Stripe, and more — each with its auth flow and the tool surface it exposes to agents.

Self-hosting is a first-class path. LLM and models walks through bring-your-own-model — Anthropic Claude, OpenAI, or self-hosted open weights via Ollama or vLLM — and how to pick per agent. Production deployment covers Docker, Kubernetes, secrets management, backups, and observability. Security is the trust model — permissions, sandboxing, credential handling, and exactly what agents can and cannot touch. Troubleshooting is the runbook for the failure modes we have hit in production.

Who each section is for

Operators and founders evaluating Maskin should start with Get started and the agentic workspace primer. Developers wiring agents into their own product can jump straight to MCP tools and the API reference. Platform engineers and SREs will get the most from Self-hosted setup, Security, and Production deployment. If you are comparing tools, the open source Jira alternative and self-hosted AI workspace articles frame Maskin against the incumbents.

Where to start

The fastest path is the quickstart tutorial — create a workspace, connect Claude over MCP, and run your first bet end-to-end in about ten minutes. From there, read core concepts to build the mental model, then pick the loop that matches your work — customer feedback, product planning, GTM, or ops — and adapt it. The MCP-native primer and the stateful orchestration substrate article give the deeper why behind the architecture. For a dated log of what has shipped recently, see the latest updates.

Articles

Prefer we run it for you? Managed hosting is available — book a meeting with the founders.