Install · Claude Code plugin
Your Claude Code work becomes verified skill signal.
The axsept plugin observes your conversations with Claude Code — what you ask, what you accept, what you push back on — and records the skills you demonstrated into your proof on axsept. Nothing leaves your machine without being anonymized first.
Before you start
- Claude Code installed. If you don't have it, grab it from claude.com/code.
- A axsept account (or you'll be prompted to create one during the install flow — the 60-day trial applies).
Install in three steps
- 1
Install the plugin
In a Claude Code session, run:
/plugin install skillchain@skillchain-marketplaceIf the marketplace isn't added yet, run
/plugin marketplace add github.com/skillchain/claude-pluginsfirst. - 2
Authorize in your browser
Your browser will open automatically to skillchain.tech/connect?source=claude-mcp. Sign in, click Authorize Claude Code, and copy the displayed
sk_user_…key. - 3
Paste the key
Back in Claude Code, paste the key when prompted for the axsept API key config. It's stored securely in your OS keychain — never in plaintext settings.
Done. Claude will start logging activities to your profile as you work — sparingly, only on substantive moments.
What gets captured
- Skills Claude observed you demonstrate (e.g., system design, debugging, code review)
- How you directed the work — whether you course-corrected, validated, or delegated
- Short evidence snippets from your own prompts, post-anonymization
- Tools and frameworks you mentioned
What doesn't
- Raw conversation transcripts
- API keys, tokens, SSH keys, JWTs (regex-stripped locally)
- Email addresses, phone numbers, SSNs, credit card numbers
- Project codenames or customer names you add to your local denylist
Local-first by design
The plugin runs as a local process on your machine. Skill extraction happens inside your Claude Code session — Claude itself observes and classifies; we don't do server-side LLM extraction. Before any data reaches axsept, it's passed through @axsept/anonymizer locally, which strips PII and secret patterns. The full pattern set is open source — you can audit exactly what we redact.
What Claude can do once installed
log_activity
Records a discrete activity you demonstrated a skill on. Called sparingly (2–8 times per hour of focused work) — not on every message.
record_session_summary
At natural session boundaries, captures your directing / orchestration skill across the whole session.
get_profile_snapshot
Read-only. Lets Claude see your current proof on axsept to avoid re-logging what's already there and to personalize coaching when you ask.
Troubleshooting
Activities aren't appearing on my profile
Claude only logs on substantive moments. If you've been asking quick questions, that's expected — try a longer working session (design, debugging, spec writing).
You can also nudge Claude explicitly: "Use the skillchain log_activity tool to record what we just did."
I need to revoke or rotate my key
/plugin config skillchain.Point the plugin at staging or self-hosted axsept
/plugin config skillchain, set the axsept base URL field to the alternate origin (e.g. https://staging.skillchain.tech). Leave blank for production.Removing the plugin
/plugin uninstall skillchain removes it from Claude Code. The key stays valid on axsept's side until you revoke it manually from Settings → API Keys — uninstall the plugin first, then revoke the key for clean rotation.Ready to connect?
If you're signed in, open the Connect page and click Authorize. If not, you'll be asked to sign in first.
The plugin is open source. Source: github.com/skillchain/integration-protocol. Protocol spec, anonymization rules, and the full attribution model are documented there.