# refresh.cv MCP

Server URL: `https://mcp.refresh.cv/mcp`

Use this remote MCP server to send completed agent work sessions to refresh.cv as editable Work History.

## Supported Agents

Use the same MCP server URL for ChatGPT, Codex, Claude, Claude Code, Cursor, Antigravity, Cline, OpenClaw, OpenCode, Hermes, Poke, and other MCP-compatible clients.

## Codex

```bash
codex mcp add refresh_cv --url https://mcp.refresh.cv/mcp
codex mcp login refresh_cv
```

## Tools

- `get_resumes`: Read the authenticated user's saved refresh.cv resumes and return resume previews. Use only for requests to find, load, inspect, or compare existing resumes or CVs. It does not create, edit, rewrite, or publish resumes and does not read Work History records. Do not call it to prepare an unsupported create, edit, rewrite, or publish request.
- `get_profile_summary`: Read a summary of the authenticated user's latest refresh.cv resume profile, including skills, experience, projects, education, and career targets. Use only when the user asks about profile context or recommendations. Do not call it to prepare an unsupported resume create, edit, rewrite, or publish request. This tool does not read Work History records.
- `check_refresh_cv_connection`: Check whether the authenticated refresh.cv connection can read resumes, Work History, and jobs. Use for an explicit connector, app, MCP, or sign-in diagnostic. Samples are omitted unless the user asks to include them.
- `get_recent_work_history`: Read the latest saved refresh.cv Work History record. Use when the user asks for recent Work History, saved activity, or recent records in refresh.cv. This returns one Work History record, not employment history from a resume.
- `list_work_history`: Compatibility alias that reads the same latest refresh.cv Work History record as get_recent_work_history. Use only when a client already targets this tool name; do not call both Work History read tools for one request.
- `save_work_history`: Save a user-approved note, research result, decision, or completed work summary as an editable refresh.cv Work History record. Use only when the user explicitly asks to save or record content in refresh.cv. The summary and structured evidence should contain only task-relevant content; do not send raw chat transcripts. A retry without the same idempotencyKey can create a duplicate record.
- `search_jobs`: Search refresh.cv job postings by role and optional location, seniority, employment type, date, company, salary, visa, or technology filters. Use for general job exploration. The service may broaden overly narrow filters when necessary and returns canonical refresh.cv job URLs.
- `match_jobs_to_profile`: Recommend and rank refresh.cv jobs against the authenticated user's latest resume profile. Use for profile, resume, skill, career, experience, or job-fit requests, including 'Use my refresh.cv profile to recommend software engineering jobs in the United States that fit my experience.' Reads profile and jobs in one call, returns fit reasons and URLs without returning the source resume, and renders its own cards. Never search the plugin directory or call render_jobs afterward.
- `fetch_job`: Fetch a refresh.cv job by ID, including its description, requirements, responsibilities, benefits, company context, and application URL. Use when the user asks for details about a job already identified in refresh.cv results.
- `get_saved_jobs`: Read the authenticated user's jobs saved or bookmarked in refresh.cv. Use for requests to view saved jobs, favorites, bookmarks, or a shortlist. This tool does not add or remove saved jobs.
- `render_jobs`: Render an explicit refresh.cv jobs array as cards only when the prior data source did not already render. Do not call after search_jobs or match_jobs_to_profile; those tools render their own cards.
- `search`: Compatibility search that returns compact refresh.cv job result IDs, titles, canonical URLs, and text for hosts that use a search/fetch retrieval pattern. Use for job-data retrieval when that host pattern is required.
- `fetch`: Compatibility fetch for an item returned by search. Retrieves refresh.cv job details and a canonical URL from a result ID such as job:<id>. Use when a host requires the search/fetch retrieval pattern.
- `get_import_status`: Check a refresh.cv Work History import, upload, or sync status by importId. Current MCP imports normally complete synchronously.

## Login Required

If an MCP request returns `error: "login_required"`, `invalid_token`, or an OAuth `WWW-Authenticate` challenge, tell the user to reconnect or authorize refresh.cv in their agent MCP/Connector settings and retry. Do not present that response as empty refresh.cv data.

If a tool result includes `structuredContent.login_required: true` or `structuredContent.reauthentication_required: true`, use `_meta["mcp/www_authenticate"]` to restart the MCP OAuth flow. Do not open `resource_metadata` or `mcp_resource` as a normal browser login URL; those values are OAuth discovery metadata.