Skip to content

[Enhancement] Ability to specify read/write base MCP Server #8

[Enhancement] Ability to specify read/write base MCP Server

[Enhancement] Ability to specify read/write base MCP Server #8

on:
issues:
types: [opened, edited, reopened]
jobs:
research:
runs-on: ubuntu-latest
if: ${{ github.event.issue.state == 'open' && contains(github.event.issue.title, 'Research') }}
steps:
- uses: austenstone/copilot-cli@main
env:
FIRECRAWL_API_KEY: ${{ secrets.FIRECRAWL_API_KEY }}
with:
github-token: ${{ secrets.PAT }}
mcp-config: |
{
"mcpServers": {
"firecrawl": {
"type": "local",
"command": "npx",
"args": ["-y", "firecrawl-mcp"],
"env": {
"FIRECRAWL_API_KEY": "${FIRECRAWL_API_KEY}"
},
"tools": ["*"]
}
}
}
prompt: |
## Role
You are a research assistant. Analyze the current GitHub issue and identify the research topic. Use Firecrawl to gather information and conduct deep research on the topic.
## Guidelines
0. **Before starting a new research task:**
- Acknowledge the issue comment by reacting with an emoji.
- Verify the issue exists
1. **For discovering information:**
- Use `firecrawl_search` when you don't know which websites contain the information
- Use `firecrawl_map` to discover URLs on a specific website before scraping
2. **For extracting content:**
- Use `firecrawl_scrape` for a single known URL
- Use `firecrawl_batch_scrape` for multiple known URLs (more efficient than multiple scrapes)
- Use `firecrawl_extract` when you need structured data (e.g., prices, names, specific details)
3. **For comprehensive coverage:**
- Use `firecrawl_crawl` carefully with appropriate `limit` and `maxDepth` to avoid token overflow
- Consider using `map` + `batch_scrape` for better control over large sites
4. **Best practices:**
- Always use reliable and up-to-date sources
- Provide summaries and key points from your research
- Cite all sources with URLs
- Use `onlyMainContent: true` to extract clean content
- Enable `deduplicateSimilarURLs` for crawls to reduce redundancy
## Input Data
```json
${{ toJson(github.event) }}
```
## Output Format
Generate a comprehensive Markdown research report and post it as a comment to: ${{ github.event.issue.html_url }}
The report should include:
- Executive summary
- Key findings with citations
- Detailed analysis sections
- Source URLs for all information