Loading…
Loading…
Loading…
For developers & power users
Works with MCPAskCV speaks the Model Context Protocol. Point Claude, ChatGPT, Cursor, or any MCP-compatible assistant at your AskCV account and it instantly has context on your experience, stories, and goals — and can push new knowledge back with your explicit consent.
Reads answer career questions in place. Writes execute immediately after an OAuth consent grant, get audited per client id, and stay reversible for 30 seconds via the undo token returned on every reversible call.
Read & import (PAT)
/api/mcpFull StreamableHTTP surface for your own assistants — query your verified knowledge base plus legacy batch-import and career-framework tools. Authenticated with a long-lived personal-access token you generate in Settings.
Read your knowledge base
Your AI assistant can surface your STAR stories, skills, and past decisions as context for career questions — no more re-typing your wins into every chat.
Search your profile
Pull role history, education, certifications, and coverage against a target role without leaving your assistant's chat window.
Check your job pipeline
Read applied / interviewing / offer statuses so your assistant can draft follow-ups with full context.
Write (OAuth PKCE)
/api/mcp/writeThe surface to hand to third-party assistants. Gated behind OAuth2 PKCE — you explicitly grant each client a scoped set of tools. Every write is audited and reversible for 30 seconds.
| Tool | Tier |
|---|---|
kb.addEntry Capture a new knowledge-base entry — e.g. a win surfaced in a 1:1 or a story extracted from a performance review. | Plus |
kb.updateEntry Refine an existing entry with new tags or updated wording. | Plus |
profile.updateHeadline Update your public headline or professional summary. | Free |
profile.updateSummary Rewrite your profile summary. Audited and undoable for 30 seconds. | Free |
profile.addExperience Append a new role with dates, description, and bullets. | Free |
pipeline.addJob Save a job posting to your pipeline with status and URL. | Free |
pipeline.updateStatus Move a pipeline job between saved / applied / interviewing / offer / rejected. | Free |
pipeline.logInterview Log an interview round with date, interviewer, and outcome notes. | Free |
coaching.logInsight Silently attach a coaching insight to an active coaching session. | Plus |
coaching.addActionItem Queue a follow-up action item surfaced during a coaching conversation. | Free |
Destructive writes (e.g. delete) stay on the first-party surface. Your external assistants can propose and capture — they cannot remove.
Connect read access in under five minutes. Upgrade to OAuth-gated writes when you’re ready to let an assistant push data back.
In your AskCV dashboard, open Settings → API Keys and create a token of type mcp. Tokens are shown once and prefixed askcv_mcp_.
Paste the snippet below into Claude Desktop’s .mcp.json, Cursor’s MCP settings, or any other MCP-compatible client config.
Claude Desktop / Cursor
{
"mcpServers": {
"askcv": {
"url": "https://askcv.ai/api/mcp",
"headers": {
"Authorization": "Bearer askcv_mcp_YOUR_TOKEN_HERE"
}
}
}
}Or fetch the discovery manifest
curl https://askcv.ai/.well-known/mcpUnauthenticated and cacheable. Advertises every endpoint, auth mode, rate limit, and the OAuth URLs your client needs.
External assistants are powerful — and external. Every AskCV control below is on by default; none of them are opt-in add-ons.
30-second undo
Every reversible write returns an undoToken. POST it back within 30 seconds to revert the change. Non-reversible writes are clearly flagged with undoToken: null.
OAuth PKCE with per-tool scopes
Writes require a PKCE-verified OAuth grant. You pick exactly which tools each client can invoke — scope format is tool:<name>. Revoking the client immediately invalidates every outstanding token.
Every write is audited
OAuth write-pack calls land in your Action Log with the originating client id, the tool name, and the resolved needsApprovalflag. MCP writes execute immediately — rely on the 30-second undo (or revoke the client) if something surprises you. Read-only tool calls are not currently logged individually.
Client-level audit trail
Every MCP write row in agent_actions carries agent_id = ‘external_agent’ plus a sourcetag pointing to the specific OAuth client id. Filter by client, export, revoke — revocation immediately invalidates every outstanding access + refresh token.
Read the full security overview or the privacy policy.
API Docs
Agent Import API
REST + MCP endpoints for importing career data. Full domain schemas, examples, and rate limits.
Read
Discovery
/.well-known/mcp
Public JSON manifest listing every MCP endpoint, OAuth URL, and rate limit. Point your client’s auto-discovery here.
Fetch
Spec
MCP spec
The canonical Model Context Protocol specification. Read it if you’re building a new client.
modelcontextprotocol.io
Sign up, generate an MCP token, and let your assistant start learning from your career in under five minutes.