MCP Wiki is a structured data connector — 28+ social networks wired as MCP tools and a REST API. Route live social signals into Claude, BigQuery, Snowflake, webhooks, or any destination your pipeline already talks to.
Strongly-typed schemas · deterministic JSON · public data only · fully auditable
Each tool returns the same response envelope — typed, versioned, ready to pipe into any destination. Structured JSON that deserialises cleanly on the other end, no translation layer, no impedance mismatch.
Velocity-weighted detection ranks breakout hashtags, sounds and formats by acceleration, not raw volume. Inflection points get flagged hours before saturation, so your agent moves while the chart everyone else watches is still flat.
Continuous surveillance of any public profile at sub-minute resolution: follower deltas, engagement decay, posting cadence, full historical lookback. When a competitor's post takes off, your agent knows in seconds and can tell you why.
Deterministic stitching links a creator across 28+ networks into a single canonical entity. Vetting an influencer stops being an afternoon of tab-hopping and becomes a single tool call returning the full cross-platform footprint.
Statistical baselining separates organic momentum from bot inflation and flags sentiment inversion the moment it breaks from the rolling norm. Delivered inline or pushed via webhook, so the bad surprise never reaches your report.
Four tiers. All 28+ social connectors available across every plan. Scale destinations, polling cadence, and call volume as your pipeline grows.
All 28+ connectors on every plan · structured JSON · cancel anytime
The brittle, rate-limited, captcha-ridden parts stay on our side. Your agent gets deterministic JSON.
One line in mcpServers and the tools appear in Claude Desktop, Cursor, Cline, Continue — anything that speaks MCP.
Every tool returns the same response envelope. Strongly-typed JSON your agent can parse without a translation layer.
Tool results come from live ingestion at source — never from a warmed cache returning yesterday's data. Your agent sees what's happening now.
Fan out hundreds of concurrent tool calls. We auto-scale the fleet behind you — your monthly quota is the only limit, never a hidden throttle.
The server reads strictly public surfaces. No credential replay, no private data, no authentication bypass — auditable end to end.
Stuck wiring up the server? Message us. You reach a person who knows MCP and the schema — not a ticket queue.
Drop one config stanza, restart your MCP client, and all four connectors are live. Pick your tier and your pipeline is in production the same afternoon.
The Model Context Protocol is an open standard for giving LLM agents tools, resources and prompts in a uniform way. An MCP server exposes capabilities; an MCP client (Claude Desktop, Cursor, Cline, Continue) connects to it and surfaces those capabilities to the model.
Anything that speaks MCP — Claude Desktop, Cursor, Cline, Continue, Zed, and rolling our own. We also expose a plain REST API for non-agent use cases, so the same key works from a backend job or a notebook.
One flat subscription per month or year with a monthly tool-call quota and feature set per tier. No per-tool multiplier and no surprise metering — the price you see is the price you pay.
No, and that's deliberate. Ingestion at this resolution is expensive to run well, and a paid-only pipeline keeps latency low and data fresh for the teams who depend on it. If you need to evaluate before committing, talk to us about a scoped pilot.
A single invocation of a tool — one discover_trends, one monitor_profile, etc. Tool definitions, prompt fetches and console usage are never metered; only outbound calls draw against quota.
We don't impose per-minute caps or artificial concurrency ceilings. Your monthly quota is the boundary — exceed it and you can upgrade or attach an overage pack without a contract change.
Never. MCP Wiki reads exclusively from publicly available surfaces. No login replay, no private endpoints, no auth bypass — an explicit compliance commitment, auditable on request.