Getting Started with MCP¶

The Model Context Protocol (MCP) is an open standard that lets AI tools — such as Claude, ChatGPT, and others — connect to external services. Content Chimera has an MCP server that gives your AI assistant direct access to your website analysis data. Instead of switching between your AI chatbot and the Chimera web interface, you can do everything from a single conversation.

Why Use MCP?¶

Work in your existing AI tool. You stay in Claude Desktop, Claude Code, ChatGPT, or whichever MCP-compatible client you prefer.
Combine Chimera data with general knowledge. Your AI assistant can pull Chimera data and reason about it alongside everything else it knows.
Automate workflows. Chain multiple Chimera operations together — crawl a site, wait for processing, run analysis, and build a report — without manual steps.
Build reports programmatically. Create charts and assemble them into reports through conversation rather than clicking through menus.
Use Chimera alongside other tools. If you have connected other MCP services (for example, a file system tool or a database), your AI assistant can use them all together in one session.

Which AI Tools Support Remote MCP Servers?¶

Any AI client that implements the Model Context Protocol can connect to Content Chimera. The following clients have been tested.

Best Bets ¶

Name	Suggestion	Cost	Modes
Claude	Generally the best for working with remote MCP servers	Requires paid plan ($20/month or more)	Web, desktop app, phone app
Le Chat (Mistral)	Good free option	Free	Web, phone app
ChatGPT	Useful if you are primarily a ChatGPT user	Free	Web, phone app
LM Studio	Excellent free and local option	Free	Desktop app

Other Options ¶

AnythingLLM — desktop (runs local LLMs)
Chatbox — desktop (runs local LLMs)
Flowise — workflow builder
Gemini CLI — command line
LibreChat — web (can be self-hosted)
Warp — terminal

Two popular apps — Copilot and Perplexity — do not currently support remote MCP servers like Chimera.

For detailed setup instructions, configuration snippets, and screenshots for each client, visit the API Tokens page in the Chimera web interface (click your name in the top-right corner, then select API Tokens).

What Can You Do?¶

The MCP server provides 17 tools organized into six categories. You do not need to memorize these — your AI assistant knows what they do and will select the right tool based on your request.

Find Your Data ¶

extent-from-url — Look up a site by URL or name and get its project ID.
capabilities-datasets — Check what processing has been completed (crawl status, databases loaded, analysis available).
data-fields — List the available data fields you can query and chart.

Ask Questions ¶

chimera-query — The main question-answering tool. Supports three modes: content analysis (structured queries about counts and metadata), semantic search (AI-powered search of actual page content), and graph database (link analysis, broken links, redirect chains).
asset-details — Get detailed information about a specific page, including metadata, screenshots, and AI summaries.

Visualize ¶

chart — Create interactive chart visualizations (bar charts, treemaps, network diagrams, and many more) or retrieve existing charts by ID.

Analyze with AI ¶

llm-fieldsets — Manage LLM prompt fieldsets that define structured AI analysis (for example, E-E-A-T scoring or content classification). Create, edit, copy templates, and test against URLs.
llm-fields — Manage individual fields within a fieldset (prompts, allowed values, result types).
summary-definition — Configure which fieldsets are used during bulk summarization.

Run Pipelines ¶

run-long-pipeline — Start a full crawl-and-process pipeline. Three variants: crawl-extended (standard crawl), sprawl (crawl plus automated Sprawl Report), and ai-readiness (crawl plus AI Readiness assessment with E-E-A-T scoring).
run-focused-pipeline — Run a single processing step on already-crawled data: extract (scraping), summarize (bulk LLM analysis), load_graph_db, or load_semantic_db.
job-status — Check the progress of running or completed jobs.

Build Reports ¶

report — Create reports and add components (text blocks, charts, images).
summaries-reports-charts — List existing summaries, reports, and charts for a site.

Other Tools ¶

annotations — Add, list, or modify annotations (notes and observations) on a site.
other-analysis — Retrieve technology detection data, logo analysis, encountered domains, and page templates.
extraction-playground — Test XPath and regex extraction patterns against a URL before running them at scale.

Example Workflow: Analyze a New Website ¶

Here is how a typical MCP session might look, described as a conversation with your AI assistant.

You: Find the Chimera project for example.com

The assistant calls extent-from-url and returns the project ID and site hierarchy.

You: What data is available for this site?

The assistant calls capabilities-datasets and tells you the crawl is complete, the flattened table is ready, and the graph database is loaded — but the semantic database has not been built yet.

You: What fields can I query?

The assistant calls data-fields and lists the available fields: URL, title, word count, folder, status code, content type, and so on.

You: How many pages are there, broken down by folder?

The assistant calls chimera-query with a structured query and returns a table showing page counts per folder.

You: Show me that as a treemap

The assistant calls chart to create a treemap visualization and returns a link to the interactive chart.

You: Are there any broken links?

The assistant calls chimera-query with a graph database query and reports the number of URLs returning 404 errors, along with the pages that link to them.

You: Create a report with the treemap and a summary of the broken links

The assistant calls report to create a new report, adds the treemap chart as a component, and adds a text block summarizing the broken link findings.

Each step takes seconds. The entire analysis — from looking up the site to a finished report — can happen in a single conversation.