Resources, Prompts, and Skills

Resources

MCP resources provide direct, read-only access to Crawlio state without calling tools. Clients can subscribe to resource updates for live data.

Static resources

URI	MIME type	Description
`crawlio://status`	`application/json`	Engine state, progress counters, seed URL
`crawlio://settings`	`application/json`	Current download settings and crawl policy
`crawlio://site-tree`	`text/plain`	Downloaded file paths as ASCII directory tree
`crawlio://enrichment`	`application/json`	All browser enrichment data across all URLs

crawlio://status tries the HTTP API first, then falls back to reading the state file on disk. crawlio://site-tree reads from file-based state. All other resources use HTTP only.

Resource template

URI template	MIME type	Description
`crawlio://enrichment/{url}`	`application/json`	Enrichment data for a specific URL

Example: crawlio://enrichment/https%3A%2F%2Fexample.com returns framework detection, network requests, console logs, and DOM snapshot for that URL.

When to use resources vs tools

Use resources when you want a live data subscription or a simple read without parameters. Use tools when you need filtering, pagination, or write operations.

Need	Use
Subscribe to crawl progress updates	`crawlio://status` resource
Get enrichment for a specific URL	`crawlio://enrichment/{url}` template
Filter logs by category and level	`get_crawl_logs` tool
Get only failed URLs	`get_failed_urls` tool
Start or stop a crawl	`start_crawl` / `stop_crawl` tools

Prompts

MCP prompts are pre-built workflow templates that AI clients can invoke. They guide the AI through multi-step operations with the right tool sequence.

Prompt	Arguments	Description
`crawl-and-analyze`	`url` (required), `maxDepth` (optional)	Crawl a site, then analyze framework, structure, and errors
`export-site`	`url` (required), `format` (required), `destination` (optional)	Crawl a site and export in a specified format
`compare-sites`	`url1` (required), `url2` (required)	Crawl two sites and compare structure, frameworks, and content
`fix-failed-urls`	(none)	Diagnose failed URLs from the current crawl and retry them

In clients that support MCP prompts (such as Claude Desktop), prompts appear in the prompt picker. Select one, fill in the arguments, and the AI receives a structured instruction to execute the workflow.

Example with crawl-and-analyze:

Prompt: crawl-and-analyze
Arguments:
  url: "https://docs.stripe.com"
  maxDepth: 3
 
The AI receives instructions to:
  1. Call start_crawl
  2. Poll get_crawl_status until completed
  3. Call trigger_capture on the seed URL
  4. Call get_enrichment to read framework data
  5. Call get_errors to check for failures
  6. Summarize findings

Skills

Skills are multi-step workflow prompts installed as files in ~/.claude/skills/. They provide guided orchestration patterns for Claude Code.

Installed skills

The crawlio-mcp init wizard installs 5 skills:

Skill	What it does
`crawlio-mcp`	Full tool reference (49 tools with parameters and examples)
`crawl-site`	Intelligent crawl workflow: start, monitor, adjust, export
`audit-site`	Multi-phase site audit: crawl, capture, enrich, analyze, report
`observe`	Query the observation log with filters
`finding`	Create evidence-backed findings from observations

One agent template is also installed:

Agent	What it does
`site-auditor`	Multi-pass analysis agent that crawls, captures, and reports

Using skills in Claude Code

Invoke a skill with the / command:

/crawl-site https://docs.stripe.com
/audit-site https://example.com
/observe host:example.com since:1h
/finding "React hydration errors on /products"

Skill distribution

Skills come from the crawlio-plugin repository. Three install paths:

crawlio-mcp init (primary): downloads from GitHub, falls back to bundled copies
npx skills add Crawlio-app/crawlio-plugin: skills.sh registry
Project-level: skills in .claude/skills/ are available when working inside the Crawlio repo

Skills are idempotent. Running init again overwrites with the latest versions.

Next steps

MCP Overview: the 3-pillar architecture and what you can do
Tool Reference: all 49 Crawlio App tools with parameters
Code Mode: 6 tools replace 49
Browser Agent: Chrome automation setup

PreviousTool Reference NextOverview