MCP CONNECTOR

Connect Claude and ChatGPT to
158,000 skills.
In a few clicks.

Skillberg exposes its proprietary graph (8 reference frameworks, 18,073 cross-standard junctions) through the Model Context Protocol. Compatible with Claude, ChatGPT and any MCP client.

Or use the custom-remote URL: https://mcp.skillberg.app/mcp
See the 15 MCP tools
● 15 tools liveOAuth 2.1 · DCR + PKCEper-user quotafr-par
claude.ai · skillberg connector
CONNECTED
U
What are the prerequisites to become a senior Data Scientist?
C
01 · Installation

Install in 30 seconds.

Two paths, depending on your client.

RECOMMENDED

Claude and ChatGPT

Add Skillberg as a custom connector: in Claude (Settings → Connectors) or in ChatGPT (Developer Mode). OAuth sign-in, zero configuration.

CUSTOM CONNECTOR · CLAUDE · CHATGPT
Skillberg
Knowledge graph · 158K skills · 8 frameworks
Connect Claude or ChatGPT
ADVANCED

Custom-remote

For Claude, ChatGPT (Developer Mode), Claude Code or any other MCP client. Set a remote URL, then follow the OAuth flow.

~/.claude/mcp.json
{
"mcpServers": {
"skillberg": {
"url": "https://mcp.skillberg.app/mcp"
}
}
}
See the install docs
02 · How to use it

Three ways to plug in Skillberg.

Same server URL, same OAuth flow — whatever your client.

SERVER URL
https://mcp.skillberg.app/mcp

Prerequisite: a Skillberg account. The OAuth flow automatically provisions an API key tied to your user.

LIVE

Claude

Claude.ai and Claude Desktop, via custom connectors.

  1. 1Settings → Connectors → Add custom connector.
  2. 2Paste the server URL https://mcp.skillberg.app/mcp.
  3. 3Sign in to Skillberg and approve the consent screen (OAuth).
  4. 4Available on plans that offer custom connectors (Pro / Team / Enterprise).
Coming soon: listing in the Anthropic connector directory (submission in progress).
DEVELOPER MODE

ChatGPT / OpenAI

Available today through Developer Mode; the public directory is on its way.

  1. 1Turn on Developer Mode in ChatGPT settings.
  2. 2Add custom connector → paste the same URL https://mcp.skillberg.app/mcp.
  3. 3Authenticate through the Skillberg OAuth flow.
  4. 4Coming soon: search for “Skillberg” in the ChatGPT App Directory (submission in progress).
HTTP / SSE

Any other MCP agent

Cursor, Cline, VS Code MCP, in-house MCP SDKs… any compatible client.

  1. 1Point the client at the server https://mcp.skillberg.app/mcp (HTTP / SSE transport).
  2. 2OAuth 2.1 auth, or an X-Skillberg-Key API key for custom integrations.
  3. 3Automatic discovery via /.well-known/oauth-authorization-server and /.well-known/oauth-protected-resource.
MCP tools docsCreate a Skillberg account
03 · Tools

Fifteen tools. One graph.

Every tool runs on the same canonical graph. URIs are stable: skillberg://skill/{uuid}, skillberg://occupation/{uuid}.

search_skillsLIVE

Semantic search across the 158,000 canonical skills.

{ query: str, limit: int = 20, threshold: float = 0.1 }
get_skillLIVE

Full multilingual skill record and its proficiency levels.

{ uri: "skillberg://skill/{uuid}" }
search_occupationLIVE

Semantic search of occupations, chainable into the profile tools.

{ query: str, limit: int = 20 }
search_sectorLIVE

Semantic search of business sectors.

{ query: str, limit: int = 20 }
skills_extractLIVE

An occupation's skill profile: essential and optional skills.

{ occupation: "skillberg://occupation/{uuid}" }
skills_recommendLIVE

Expands a set of skills: prerequisites and related skills.

{ skills: ["skillberg://skill/{uuid}", …] }
skills_gapLIVE

Gap between a candidate and a target occupation: covered, missing, coverage ratio.

{ candidate_skills: [uri, …], target_occupation: uri }
career_transitionsLIVE

Occupations reachable from an origin occupation, ranked by shared skills.

{ from_occupation: "skillberg://occupation/{uuid}" }
enrich_cvLIVE

Maps a CV's skills (JSON Resume) onto the frameworks, no prior upload.

{ resume: JSONResume }
evaluate_cvLIVE

CV quality and completeness: overall score and per-section breakdown.

{ resume: JSONResume, section_weights?: object }
enrich_jobLIVE

Extracts and normalises the required skills of a job offer.

{ jsonJob: object }
match_cv_to_jobsLIVE

Fit score between a user's uploaded CV and a job.

{ user_id: str, job_id: str }
render_skills_radarLIVE

Radar chart comparing several profiles' skills, rendered as an image.

{ profiles: [...], title?: str, scale_max?: number }
render_tableLIVE

Table of candidates, offers or a summary, ready to share.

{ kind: "candidates" | "offers" | "summary", rows: [...] }
credits_remainingLIVE

The key's credit balance: remaining, monthly allowance, usage.

{}
15 tools live · all read-onlyView the OpenAPI spec
04 · Conversations

Three concrete examples.

Real tools; the answers are illustrative.

Skills mapping for a job
tools: search_occupation + skills_extract
U
Find the key skills of a senior tech Product Manager.
C
Checking the Skillberg graph…
search_occupation(query="Senior Product Manager")
skills_extract(occupation="skillberg://occupation/…")
Product DiscoveryESCO
RoadmappingESCO
Stakeholder ManagementO*NET
Data-informed prioritizationESCO
Technical architecture literacyROME
Skill gap to a target occupation
tools: enrich_cv + skills_gap
U
What does this CV lack for a DevOps engineer role? [paste CV]
C
enrich_cv(resume=…)
skills_gap(candidate_skills=[…], target_occupation=…)
Coverage
78%
Covered12 / 15MissingKubernetes, Terraform
No prior upload: the CV is passed directly.
Career transitions
tools: search_occupation + career_transitions
U
Which roles can a maintenance technician move into?
C
career_transitions(from_occupation="skillberg://occupation/…")

Closest occupations, by share of shared skills:

Methods technician64%
Automation technician58%
Maintenance manager51%
Ranked by skills shared with the origin occupation.
05 · Auth

Authentication

OAuth 2.1 with Dynamic Client Registration (DCR) and PKCE. No shared passwords. Every user (Claude, ChatGPT…) gets their own Skillberg key, provisioned automatically on first sign-in.

  • One key per user (not per organization).
  • Quota and usage visible in console.skillberg.app.
  • Instant revocation, from Skillberg or from the client (Claude, ChatGPT).
06 · Privacy

Privacy

Text sent through the tools (CVs, job descriptions) is processed on the fly and never stored. Graph queries are logged for quota and debugging — not for training.

  • Inference in France (Scaleway fr-par).
  • European models (Mistral).
  • GDPR + AI Act ready.
07 · Use cases

Who it's for.

HR and internal mobility

Your Claude or ChatGPT agents answer HR questions from a canonical graph instead of hallucinating.

Learning and training

Build learning paths on validated prerequisites (Skill Trees).

Strategy and planning

Skill-gap analysis at scale, with numbers you can defend.

Pricing
Free to try. Pay-as-you-go after that.

Discovery tier included with every Skillberg account. 30-day free trial on your first MCP connection. Pricing based on the number of calls beyond the quota.

See pricing
08 · Get started

One connection. Fifteen tools.
One hundred fifty-eight thousand skills.

Thirty seconds to connect Skillberg to Claude or ChatGPT.

Connect Claude or ChatGPT

Or: custom-remote URL https://mcp.skillberg.app/mcp

Need help integrating? contact@skillberg.app