Base44
Build and edit Base44 apps from an AI assistant, and query data your app's users submit.
Query your IcePanel architecture model and decision records from an AI agent.
Opens Agent Studio, where connecting is one click. The connector URL below works in any MCP client.
Paste it into any MCP client. Setup docs
The IcePanel MCP server connects a software architecture model to Claude and any MCP-compatible agent. Its 24 tools read model objects, connections, diagrams, flows, teams, technologies and decision records — and can create or update objects, connections and ADRs, if an admin has enabled write access.
Connection verified
Connection checked by Agentman on .
Listed in Anthropic’s Connectors Directory at community tier — a listing, not a partner-level review.
Anthropic states this reflects the level of review a connector received, not a security audit.
Tool names from Anthropic's directory listing. This server requires sign-in, so we could not read tool descriptions or parameter schemas.
docs.google.com document rather than a page on its own domain, which is worth knowing if you need a stable citation for review.Only if an admin has turned that on. IcePanel documents that all editors and admins can set up MCP with read access by default, but write access is disabled by default for existing organizations and must be enabled by toggling OAuth write access on the AI settings page. The scopes reflect this split: landscape.read and landscape.write.
A paid one. IcePanel states the MCP server is only available on paid plans, with no free-tier access documented. Beyond the plan, access also depends on your role: editors and admins can configure MCP with read access, and write access requires the separate organization-level toggle described in the AI settings page.
IcePanel publishes an explicit unsupported list: creating diagrams, draft tools, version tools and comment tools. So an agent can add model objects, connections and decision records, but cannot create a diagram, work with drafts or versions, or post comments. That published negative is unusually clear for a connector.
IcePanel's own examples include what your landscapes are, details about any model object, what an API service is connected to, which technologies are most used in your system, what a given team owns, and whether an architecture reflects your decision records. The last one pairs the model with ADRs, which is the connector's most distinctive read.
Yes, for both reading and writing. The tool set includes listing ADRs and getting ADR details, plus creating and updating them. IcePanel names decision records in both its supported read tools and its write and delete tools, so an agent can check whether a design matches recorded decisions and then record a new one.
IcePanel names Claude, Claude Code, Microsoft Copilot Studio, Codex, Cursor and VS Code, and tells you to look for Custom MCP connectors support in your tool. Any client that supports remote MCP over OAuth should connect, but those six are the ones IcePanel says it has actually tested.
.md variant; retrieved 2026-09-13), source of the paid-plan gate, the supported and unsupported tool lists, the tested-client list and the write-access toggle. docs.icepanel.io/robots.txt carries Content-Signal: ai-train=yes, search=yes, ai-input=yes — an explicit publisher opt-in to AI synthesis, recorded here because we rely on it. · retrieved 2026-09-13https://mcp.icepanel.io/mcp — HTTP 401 with a www-authenticate Bearer challenge carrying realm="mcp", and RFC 9728 metadata at https://mcp.icepanel.io/.well-known/oauth-protected-resource/mcp returning scopes_supported of landscape.read and landscape.write, resource_name "IcePanel MCP Server", and https://api.icepanel.io as the authorization server (2026-09-13) · retrieved 2026-09-13Connect IcePanel once and your agents call these tools on their own: on a schedule, in a workflow, with nobody at the keyboard.