Connect your AI to your skills vault

SkilzVolt is a remote MCP server. Claude, ChatGPT, Claude Code, Codex — and any other MCP client — connect with nothing but a URL: sign-in, permissions, and registration all happen automatically through OAuth.

Claude (web, desktop & mobile)

In Claude, open Settings → Connectors → Add custom connector.
Name it SkilzVolt and paste the server URL: https://skilzvolt.com/mcp. Leave the OAuth Client ID/Secret fields empty — Claude registers itself automatically.
Click Add, then Connect. A SkilzVolt window opens: sign in with Google or Microsoft, choose the access level, and approve.

That's it. Claude can now search, read, and propose changes to the skills in every workspace you belong to — including workspaces you join later. No keys to copy, nothing to rotate by hand.

ChatGPT

In ChatGPT, open Settings → Apps → Advanced settings and enable Developer mode (required for custom MCP apps).
Click Create app: name it SkilzVolt, set the MCP Server URL to https://skilzvolt.com/mcp, and choose OAuth as the authentication method. Advanced OAuth settings can stay on their defaults — SkilzVolt supports Dynamic Client Registration, so ChatGPT discovers everything automatically.
Connect: sign in with Google or Microsoft, choose the access level, and approve.

SkilzVolt also implements ChatGPT's standard search and fetch connector tools, so chat search and deep research can pull from your vault too.

Claude Code, Codex & other MCP clients

Any client that supports remote MCP servers with OAuth works the same way.
Claude Code: claude mcp add --transport http skilzvolt https://skilzvolt.com/mcp then /mcp to authenticate.
Codex CLI: codex mcp add skilzvolt --url https://skilzvolt.com/mcp then codex mcp login skilzvolt to run the OAuth sign-in.
Clients without OAuth support can use a pasted agent key instead — create one from your workspace's Agents page.

OpenClaw, Cursor, Windsurf & anything else that speaks MCP

SkilzVolt is a standard remote MCP server — nothing about it is vendor-specific. Most agents accept the common mcpServers config block:

{
  "mcpServers": {
    "skilzvolt": {
      "url": "https://skilzvolt.com/mcp"
    }
  }
}

If the agent supports MCP OAuth it will walk you through sign-in automatically. If it only supports headers or keyed URLs, create an agent key on your Agents page and use https://skilzvolt.com/mcp/svk_your_key or Authorization: Bearer svk_your_key — same tools, same permissions, revocable any time.

What your agent can do

Ten tools are read-only; three propose changes. Nothing an agent does bypasses your review gates: writes create proposals with word-level diffs, routed by each skill's approval mode (automatic, agent self-review with audited sign-off, or human approval). Agents never destroy data — versions are immutable and history is never rewritten.

search	read	ChatGPT-standard connector search across your skills (id, title, url results)
fetch	read	ChatGPT-standard fetch of one skill's full content by id
workspaces_list	read	List the workspaces the connection can access and the access level in each
skills_search	read	Search skills by name and description across accessible workspaces
skills_get	read	Read a skill's current content and resource list
skills_get_resource	read	Fetch one resource (reference, script, template, data) attached to a skill
skills_export_all	read	Paginated export of all current skills including governance metadata
skills_check_pending	read	List a skill's open change proposals
skills_proposal_status	read	Check the status and outcome of one proposal
skills_get_proposal_diff	read	Fetch a proposal's word-level diff and review package
skills_create	write	Propose a new skill (goes through the workspace's review gate)
skills_propose_change	write	Propose an edit to an existing skill (never overwrites directly)
skills_submit_review	write	Sign off a proposal pending agentic review, with a mandatory audited rationale

Security & revocation

OAuth 2.1 by default — dynamic client registration, PKCE (S256), short-lived access tokens with rotating refresh tokens. Tokens are stored hashed; authorization codes are single-use.
Least privilege — you pick read-only or read & propose at connect time. Workspace access follows your memberships live: leave a workspace and the connection loses it immediately.
One-click revocation — every connection appears on your Agents page and can be revoked at any time; revocation takes effect on the next request.
Audit trail — every proposed change, review decision, and connection event is recorded and attributed to the connection and its owning user.

Need help?

Questions or issues connecting? Email hello@skilzvolt.com — we respond quickly. See also our Privacy & AI policy.