mcp
SkillProductivityUse MCP (Model Context Protocol) servers from bash with `shelley mcp`. Use when the user mentions MCP, asks to add or connect an MCP server, or a task needs an external service that may be reachable through a registered MCP server.
Use mcp in Claude, ChatGPT or Ahel Desktop
Free. Sign in, add mcp and connect your AI. About a minute.
Also: Claude Code · Cursor · Codex
Then ask your AI: use the mcp skill
Details
Instructions available. Your AI can read the instructions. Execution depends on the setup they require.
Account requirements not reviewed. Check the skill instructions before use; ahel provides instructions and does not run this skill.
No other account needed.
Add ahel to your AI once: Claude, ChatGPT, Cursor, Claude Code or Codex. Then ask it to use this.
What this skill tells your AI
The instructions your AI receives, as published by boldsoftware/shelley in skills/builtin/mcp/SKILL.md and read by ahel’s review.
MCP servers registered with Shelley aren't model tools: use them by running shelley mcp from bash (shelley mcp -h for all commands).
shelley mcp list # servers
shelley mcp list SERVER # its instructions and tools as TypeScript-style signatures
shelley mcp list SERVER.TOOL -schema # one tool with its JSON schemas
shelley mcp search issue label # tools matching every term, across all servers
shelley mcp search SERVER issue # ...limited to one server
shelley mcp call linear.search_issues query='login bug' limit=5 'states=["open"]'
To find a tool, prefer shelley mcp search TERM... over grepping list: it
matches the terms against each tool's name, description and parameters and
prints the full documentation of every matching tool, grouped by server.
Read a server's instructions and signatures before first calling its tools. KEY=VALUE values are taken as is for string parameters and as JSON otherwise, so quote JSON for the shell. Images, audio and binary resources are saved to temporary files (the output says where; look at images with read_image). If the tool reports an error, it is printed to stderr and the exit status is 1. A call that timed out may still have run.
Compose calls like any shell command; keep output small with jq:
shelley mcp call SERVER.TOOL -json | jq '.structuredContent.items[] | {id, title}'
jq -n --arg q "$query" '{query: $q, limit: 5}' | shelley mcp call SERVER.search - # arguments as JSON on stdin
for id in 1 2 3; do shelley mcp call SERVER.get_item id="$id"; done
shelley mcp call A.list_rows -json > /tmp/rows.json
jq -c '.structuredContent.rows[] | {title: .name}' /tmp/rows.json |
while read -r args; do echo "$args" | shelley mcp call B.create_item -; done
Log in
When a command says a server needs you to log in, give the user the link exactly as printed, wait for them to say they've logged in, and retry. Never open or fetch the link yourself: it's for the user's browser, and only they can log in. shelley mcp auth SERVER shows the login state and link. Never ask for or invent tokens for servers that log in this way.
Register
shelley mcp add NAME -d 'what it is for' https://example.com/mcp # logs in with OAuth if needed
shelley mcp add NAME -H 'Authorization: Bearer TOKEN' https://example.com/mcp
shelley mcp rm NAME
For servers that take an API key, ask the user for it. Shelley only connects to Streamable HTTP servers, not to stdio ones (local commands such as npx ...).
Editing a server's URL to tweak its path or query (a common way to pick which tools it exposes) keeps an existing OAuth login; only changing the scheme or host, or giving it an Authorization header, drops it.
Sessions
Each conversation has its own session with each server, so server state (open pages, selections) lasts across calls; forks and subagents start fresh. shelley mcp restart SERVER starts this conversation's session over, losing its state.
Signals
- GitHub stars
- 676
- Forks
- 111
- Last commit
- Oct 2026
Advanced
- Item type
- skill
- Key
mcp-boldsoftware- Source
- github.com/boldsoftware/shelley
github.com/boldsoftware/shelley
Related picks
Skill · mattpocock
The pick for TypeScripttypescript-pro
Skill · jeffallan
The pick for TypeScriptsocial
Skill · coreyhaines31
More in Productivitygws-calendar
Skill · googleworkspace
More in Productivitylark-minutes
Skill · larksuite
More in Productivitylark-workflow-standup-report
Skill · larksuite
More in Productivity