Rolli MCP

What the Rolli MCP MCP server does

The Rolli MCP MCP server exposes Rolli IQ’s social search and intelligence features through MCP. It can search for keywords or hashtags, inspect user profiles, identify experts from natural-language requests, and retrieve the resulting data. Supported source platforms listed by the project include X, Reddit, YouTube, Facebook, Instagram, Threads, and Bluesky.

Search results can include status information, analytics, posts, profile metrics, content analysis, and AI-generated explanations for why recommended experts match a request. Keyword searches also support conversation topic trees and raw-post retrieval. User and keyword searches can be listed and filtered by status.

The server also provides access to Rolli Agent investigations. An agent can start an investigation and receive a Markdown report, inspect progress or completed results, retry failed or stuck runs, cancel active work, rename or pin runs, and soft-delete completed or failed runs. Scheduled reports can be created for daily, weekly, monthly, or one-time execution.

How it works

MCP tools send requests to Rolli IQ using an API token. Search and investigation operations return results when complete, while the server can wait for asynchronous work using configurable polling intervals and a maximum wait time. Agents can separately retrieve the status or results of existing searches and runs by ID.

Schedules expose their recent executions, and a separate history tool lists the full execution record for a schedule. Integration tools read or change the configured webhook URL and integration name, allowing search-completion notifications to be directed to an external endpoint. Usage data includes API consumption and per-user breakdowns.

Setup and configuration

The Rolli MCP MCP server requires a Rolli account with API access and a token obtained from Rolli. The documented setup runs the published @rolli/mcp package with npx and passes configuration through the MCP client’s server definition. Smithery deployment is also supported through the repository’s smithery.yaml, with the token requested during setup.

ROLLI_API_TOKEN is required. ROLLI_USER_ID is optional and defaults to rolli-mcp; use it when requests should carry a different user ID. ROLLI_POLL_INTERVAL_MS controls the wait interval, defaulting to 5,000 milliseconds, while ROLLI_MAX_POLL_MS sets the maximum wait period and defaults to 600,000 milliseconds.

The README provides configurations for Claude Desktop, Claude Code, VS Code, Cursor, and Windsurf. Local development uses npm commands for dependency installation, linting, TypeScript builds, and tests.

Tools and capabilities

The tool groups cover:

• Keyword, user, and expert search creation, listing, and result retrieval
• Raw post access and topic-tree analysis
• Rolli Agent investigations and lifecycle management
• Recurring report creation, updates, pausing, resuming, and deletion
• Webhook configuration and API usage reporting

Limitations and notes

The provided material documents search, analysis, investigation, scheduling, webhook, and usage operations. It does not describe tools for publishing posts, replying, following accounts, or managing social-media credentials. Search and investigation completion can take time, and the configured maximum polling duration determines when the server stops waiting.

Rolli MCP MCP server credentials should be configured through the MCP client environment. If an external workflow supplies social-post context to an investigation, the repository guidance says to keep credentials separate and pass only reviewed source details such as handles, URLs, IDs, excerpts, or notes.