# Agent Coach (Upskill) [Health: Active]

**Category:** 💻 Developer Tools  
**Repository:** https://upskill.coach/  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/agent-coach-upskill

## Description
Agent Coach is a free coach for AI agents. Search a library of agent skills, MCP servers and prompts that passed a security scan (prompt injection, malicious code, secret leaks) at a pinned commit, get recommendations matched to an agent's goal, and run daily check-ins that say what to keep, mark temporary or remove. Remote MCP server with 13 tools plus a REST API (OpenAPI). Operated by Avatar 8 LLC.

## Tools
Capabilities this server exposes over MCP:

- **register_agent** — Use this first, once. Creates your Agent Coach profile and returns agent_id and api_key (shown once: save it in the owner's sanctioned secret store or a restricted credential file, not in chat or memory). Also returns example north-star metrics and suggested KPIs for your job type.
- **set_goals** — Use this after registering (and whenever goals change). Goal intake: job_type, stage, ONE north-star metric with a fixed definition and baseline, the named deliverable, capability gaps, goals, optional secondary KPIs, tools, constraints. Changing the north-star definition is recorded as drift.
- **get_recommendations** — Use this after set_goals. Returns up to `limit` catalog items matched to your goals (plus at most one extra per declared capability gap), each with reasons, its vetting status and basis, cautions, and whether it needs owner approval. Show them to your owner before installing anything.
- **update_items** — Use this after your owner approves an item (action=adopt), rejects one (decline), when an item doesn't fit your job (not_relevant: not a judgment on its quality), or confirms removing a pending_removal item (remove). The coach measures only adopted items.
- **check_in** — Use this once a day. Report the north-star value, secondary KPIs, releases (shipped vs verified live), build vs distribution actions, owner blockers, repeated corrections, timeouts, unsupported claims, incidents, deliverable progress, and per-item usage. Value format everywhere: {value, evidence: self_reported|verified, source} (a bare number is accepted and treated as self_reported). Returns health, signals, item verdicts and actions.
- **get_report** — Use this to see your KPIs vs baseline, every item's verdict with evidence, and check-in history.
- **share_report** — Use this only with your owner's OK. public=true returns a public URL (display name, goals, KPIs, verdicts; never email, notes or keys). public=false hides it.
- **get_digest** — Use this to preview the daily owner digest (subject, html, text). This call sends nothing; the daily email goes out only if owner_digest was turned on.
- **rotate_key** — Use this if your key may have leaked. Returns a new api_key; the old one stops working immediately. Save the new one.
- **delete_agent** — Use this only when your owner asks. Permanently deletes your Agent Coach profile, items, check-ins and public report. confirm must equal your agent_id.
- **get_rules** — Use this to see exactly how the coach decides item verdicts (permanent / promising / temporary / trial / unproven / pending_removal), signals, and overall health.
- **search_library** — Use this to find skills, MCP servers, prompts or repos by keyword outside your matched recommendations. Returns only items that passed the security scan, each with install, license, source and a page link. Catalog data only: no user comments. Show results to your owner before installing.
- **get_library_item** — Use this to fetch one vetted library item by slug: install, source, license and (when the license allows) the full SKILL.md text. Counts as a download. Unvetted items are not returned.

## Claude Desktop Quick Installation
No generated config for this server: it installs through its own setup steps.

The project's own setup steps (from https://upskill.coach/docs/mcp), copied verbatim. Run them in order; they are not an mcpServers config.

1. Add to Claude Code:

```sh
claude mcp add --transport http agent-coach https://upskill.coach/mcp
```

## Documentation

## What Agent Coach (Upskill) does

Agent Coach (Upskill) gives an AI agent a profile, a structured goal set, and an ongoing record of progress. It is operated by Avatar 8 LLC and is available as a remote MCP server alongside a REST API described with OpenAPI.

The service focuses on agent improvement through a vetted library of skills, MCP servers, prompts, and repositories. Catalog entries have passed security checks for prompt injection, malicious code, and secret leakage at a pinned commit. Recommendations are matched to the agent's stated job, stage, north-star metric, deliverable, capability gaps, constraints, and available tools.

Adoption is owner-controlled. An agent can find and evaluate catalog entries, but recommendations and search results are intended to be shown to the owner before installation. The service measures adopted items rather than items merely discovered or recommended.

## How it works

Registration is the first operation. `register_agent` creates an agent profile and returns an `agent_id` and an API key; the key is shown once and should be stored in an owner-approved secret store or restricted credential file. The response also includes example north-star metrics and job-specific KPI suggestions.

`set_goals` records the agent's job type, stage, one precisely defined north-star metric and baseline, named deliverable, capability gaps, goals, optional secondary KPIs, tools, and constraints. A change to the north-star definition is treated as drift.

After goals are set, `get_recommendations` returns matched catalog items, reasons for each match, vetting status and basis, cautions, and whether owner approval is needed. `update_items` records adoption, rejection, irrelevance, or confirmed removal. Agents can then use `check_in` daily to report metrics, evidence sources, releases, build and distribution work, blockers, corrections, timeouts, unsupported claims, incidents, deliverable progress, and per-item usage.

## Tools and capabilities

The 13 MCP tools cover the complete coaching loop:

- Register an agent, define goals, and retrieve matched recommendations.
- Adopt, decline, mark as not relevant, or remove catalog items.
- Submit daily check-ins and review KPI reports, item verdicts, and history.
- Preview the owner digest and optionally create a public report URL with owner approval.
- Rotate a potentially exposed API key or permanently delete an agent profile.
- Inspect the verdict rules and search or retrieve vetted library entries.

`search_library` supports keyword searches outside the recommendation set. `get_library_item` returns installation details, source, license, and, when permitted by the license, the complete `SKILL.md` content. Retrieval counts as a download. The REST API provides an additional OpenAPI-described interface.

## Limitations and notes

Security scanning applies to catalog entries at a pinned commit; it does not make arbitrary external items eligible for recommendation or retrieval. Unvetted items are not returned by the library tools.

Public reporting requires an explicit sharing action. A public report can include the display name, goals, KPIs, verdicts, and a public URL, while excluding email addresses, notes, and keys. `get_digest` only previews the daily owner digest; delivery depends on the owner digest setting.

API keys require careful handling because registration displays the original key only once. If a key may have leaked, `rotate_key` immediately invalidates the old key. Deleting an agent permanently removes its profile, items, check-ins, and public report, and requires confirmation using the agent ID.

_Full upstream README: https://allmcps.com/mcp/agent-coach-upskill/readme_