Swarmix is an MCP server that gives your AI coding agents structure, memory, and governance. It runs alongside your editor (VS Code, Cursor, etc.) and provides conversations, a knowledge base, development plans, tickets, and decision tracking — all accessible to your agents via the MCP protocol.

No. Swarmix does not include or run any AI models. It sits between you and your agents as a coordination layer. You keep using Claude Code, Cursor, Copilot, or whatever tools you already have. Swarmix augments the process around them — persistent memory, structured decisions, gated plans — so your agents work within a process instead of guessing every session.

You add a Swarmix MCP configuration to your editor (a .mcp.json file) and start the server. Your agent connects on the next session and has access to the full set of coordination tools. The knowledge base, conversations, and plans persist between sessions.

Any editor or tool that supports the Model Context Protocol (MCP). That includes VS Code with Claude Code, Cursor, and any MCP-compatible client. The agent framework doesn't matter — Claude Code, LangChain, CrewAI, or anything that speaks MCP.

Decisions, knowledge base entries, open plans, tickets, and conversation history all live on the server and persist between sessions. You decide what goes into the knowledge base — at the end of a session you tell the agent to save what matters. At the start of the next session, you tell the agent to read the relevant entries. The context is there when you need it, but you control what gets stored and what gets loaded.

The knowledge base stores versioned documents — architecture decisions, constraints, reference material, anything your agents should know. Every change is tracked with a version history and the author who made it. Agents read the knowledge base on connect and reference it during conversations.

Gates are checkpoints between plan phases. They define conditions that must be met before work continues — tests passing, a security review, a proposal being accepted. The agent can't skip ahead to the next phase until the gate is satisfied. It's how you keep speed from creating debt.

No. You direct every action. Agents can propose, review, and vote, but you decide when to submit a proposal, who reviews it, and when to close it. Autonomous systems are not our goal — you're at the centre and you lead the way.

Yes. 14 days, no credit card required. You get all core features — conversations, knowledge base, plans, governance, proposals. The trial is limited to 1 member, 1 project, 3 agents, and 10 conversations per month.

Individual is for a solo dev working with their own agents. Team is for when you bring colleagues in — more members, more projects, more agents, and access to all plugins including Jira, Linear, Slack, and webhook integrations.

Yes. You can upgrade or downgrade at any time. Upgrades take effect immediately. Downgrades take effect at the end of your current billing period. If you downgrade, plugins that require a higher tier will be suspended.

Your data is stored in Zurich, Switzerland. We encrypt data in transit (TLS) and at rest (AES-256), and we're compliant with Swiss data protection law (FADP) and the EU GDPR. We do not send your data to any third-party AI services — Swarmix is a coordination layer, not an AI provider. Your conversations, decisions, and knowledge base stay within our infrastructure.

Not yet. We understand the concern — some teams need their data on their own infrastructure. We're evaluating how to offer an on-premises installation, but updates, maintenance, and support for self-hosted deployments add real complexity. We're looking into the best way to deliver this without compromising the experience. If this matters to you, get in touch. In the meantime, see our dedicated hosting option below.

Yes. Under the Uncapped plan we offer dedicated hosted environments — a completely separate infrastructure spun up for your organisation, managed and maintained by us. Your data stays isolated in its own architecture while we handle security, updates, and operations. Same compliance guarantees, no shared tenancy. Contact us to discuss your requirements.

No. Swarmix does not include or operate any large language models. Your conversation data, proposals, knowledge base entries, and all other content are used solely to provide the coordination service. Nothing is sent to any AI training pipeline.

Your content is available for export for 30 days after cancellation. After that, it's permanently deleted. Billing records are retained for 10 years as required by Swiss law.

Yes. Bidirectional sync is available via plugins. Tickets created by your agents appear in your issue tracker, and changes made in the tracker sync back to Swarmix. Status mapping, priority, and assignees are all configurable. Jira and Linear require the Team tier; ClickUp and GitHub are available from Individual.

Yes. We have adapter plugins for both frameworks. Your LangChain agents get Swarmix tools as native LangChain tools, and CrewAI crews get BaseTool wrappers. Both require the SDK Access plugin (Team tier).