MCP & IDE agents
Run Piloterr inside Hermes Agent with config.yaml
Hermes Agent (Nous Research) keeps MCP servers in ~/.hermes/config.yaml under mcp_servers. Set url and headers.x-api-key, then /reload-mcp. The same memory can call the library from the desktop app, CLI, or chat surfaces Hermes already supports.
- YAML home config: ~/.hermes/config.yaml, then /reload-mcp
- Persistent memory and scheduled jobs, not a one-shot IDE chat
- Hosted mcp.piloterr.com, full Piloterr library as tools
- Fits Telegram, Discord, and other Hermes surfaces once MCP is loaded
At a glance
YAML
config.yaml
/reload-mcp
apply config
190+
library tools
Memory
stays across chats
Why connect Hermes Agent
Reload is part of the product
Editing YAML is not enough. Hermes needs /reload-mcp (or a restart) before new servers appear. Cursor applies .mcp.json on its own schedule.
mcp_servers, underscore, YAML
The key is mcp_servers, not mcpServers. JSON from Cursor will not parse as YAML without a rewrite.
Built for unattended work
Hermes can schedule briefings. Point those jobs at Piloterr crawl or news tools instead of a fragile local browser.
One memory, many surfaces
After reload, the same tool list is available wherever that Hermes instance runs. You do not add a second connector per app.
Hermes + Piloterr that IDEs do not cover
Long-running agent, not an editor plugin.
Morning SERP brief
Schedule Google News or search tools, write a note into Hermes memory, skip the IDE.
Watch a competitor URL
Crawl or render on a cron-like Hermes task, then ping you on Telegram when the JSON changes.
Skills that remember selectors
Hermes can store how a scrape was solved. Next run, it calls the same Piloterr tool with the same args.
Desktop, CLI, and chat
Install path is still YAML plus reload. The UI you use after that is Hermes, not ChatGPT settings.
Hermes Agent vs a session-based IDE agent
| Approach | DIY | Piloterr |
|---|---|---|
| Config format | JSON or TOML in other clients | YAML mcp_servers |
| Apply changes | Restart Cursor / Codex | /reload-mcp |
| Memory | Mostly one chat | Persistent Hermes memory |
| Surfaces | One IDE window | Desktop, CLI, chat apps |
Add Piloterr to Hermes Agent
Step 1
Create an API key
From the Piloterr dashboard. It becomes headers.x-api-key.
Create an API keyStep 2
Edit ~/.hermes/config.yaml
Under mcp_servers, add a piloterr entry with url https://mcp.piloterr.com/ and the header.
Step 3
Run /reload-mcp
In Hermes, reload MCP so the process picks up YAML. A file save alone will not attach the server.
Step 4
Ask for a library tool
Try a crawl or a marketplace lookup. If the list is empty, confirm YAML indentation and /health.
Workflow recipes
Nightly marketplace scan
Hermes calls a search tool on a schedule and stores new SKUs in memory.
Telegram “price this URL”
You drop a product link in chat; Hermes uses the matching Piloterr product tool.
Weekly jobs digest
Indeed or similar tools, then a short memo Hermes files for Monday.
Keep a render recipe
Once WebUnlocker works on a host, Hermes remembers the tool choice for the next ticket.
Hermes vs other Piloterr MCP clients
Scenario
You already run Hermes Agent
Recommendation: This page, YAML + reload
Scenario
You want a coding IDE
Recommendation: Cursor or Claude Code
Scenario
You want ChatGPT’s UI
Recommendation: Developer mode connector
Scenario
You want batch automation
Recommendation: n8n or the SDK
Hermes YAML snippet
Home file. After you save, run /reload-mcp. This is not Cursor JSON.
mcp_servers:
piloterr:
url: "https://mcp.piloterr.com/"
headers:
x-api-key: "your-x-api-key"See also
MCP
Cursor
ChatGPT
Claude CodeTransparent credit pricing
Pay only for successful requests. Start with +500 credits, then scale with plans from $49/mo.
Premium
$49/mo
18,000 credits
Premium+
$99/mo
40,000 credits
Startup
$249/mo
110,000 credits
Ready to get started?
Your web scraping API is one click away. Start with +500 credits, no infrastructure to set up, no proxies to manage, and no credit card required.
- +500 credits
- No credit card required
- All endpoints included