Wisepanel MCP

What ikoskela/wisepanel-mcp MCP server does

The ikoskela/wisepanel-mcp MCP server gives an MCP-compatible client access to Wisepanel deliberations. A deliberation takes a question, generates roles around the relevant decision space, and asks multiple AI agents to examine it from different perspectives. The default panel spans models from Anthropic, OpenAI, Google, and Perplexity, according to the README.

The intended use is not routine question answering. It is aimed at decisions where an overlooked assumption, unresolved objection, or alternative trade-off could materially affect the outcome. Examples in the project documentation include architecture choices, migrations, security and privacy trade-offs, and vendor selection.

How it works

Before starting a panel, wisepanel_magic_prompt can rewrite loaded framing, embedded assumptions, or false binaries. The resulting question is then used to create roles that each hold a different tension, such as speed versus safety or cost versus access. Agents are expected to work through both sides of their assigned conflict rather than simply advocate for one position.

The panel topology places agents across connected discussion points. An agent can carry conclusions from one discussion into another, allowing observations from separate conversations to circulate through the panel. Speaking order is balanced so that one participant does not consistently establish the initial frame or provide the final response.

Runs are asynchronous. Starting a deliberation returns a run_id; polling can wait for new events and expose responses as they arrive. The full result can be retrieved after completion if live polling was not used.

Setup and configuration

The ikoskela/wisepanel-mcp MCP server runs locally as a stdio process rather than as a hosted HTTP MCP endpoint. Install or launch it with:

npx -y wisepanel-mcp

Set WISEPANEL_API_KEY to the key obtained from Wisepanel. The value should be the complete wp_sk_... string without a Bearer prefix; the server adds the authorization scheme itself. WISEPANEL_API_URL is optional and defaults to https://api.wisepanel.ai.

For Claude Code, the documented configuration passes the environment variable and launches the package through npx. Other MCP clients can use the same command and environment-based configuration. A successful process connection does not validate the key; authentication is checked when a deliberation call is made.

Tools and capabilities

• wisepanel_start starts a panel and returns its run ID.
• wisepanel_magic_prompt revises question framing before deliberation.
• wisepanel_poll long-polls a running run for up to 15 seconds and returns arriving panelist responses.
• wisepanel_result retrieves the complete output of a finished run.
• wisepanel_cancel stops an active deliberation.
• wisepanel_publish makes a completed deliberation publicly viewable through Wisepanel Commons.
• wisepanel_list_runs lists deliberation runs in the current session.

The README also states that reasoning can be audited by default and that optional native web search can check dates, citations, figures, and rules. These options are associated with deliberation configuration rather than separate MCP tools in the provided listing.

Limitations and notes

A panel is slower and costs more than a single query, so it is not intended for questions with established answers or situations where dissent would not change a decision. The server requires a Wisepanel API key, and the provided material does not specify service pricing or an open-source license. Publishing is a deliberate visibility change: completed results sent to Wisepanel Commons become publicly viewable and shareable.