Conference Partner — MCP server & REST API

Academic conference and journal data for AI agents and scripts: CFP deadlines, CCF / CORE / QUALIS rankings, the Conference Partner Index (CP-I) — a 0–100 score we compute for every conference, including the majority that carry no external rank — acceptance-rate history, edition history, journal impact factors and special issues — served by Conference Partner (会伴), a conference tracker that has been maintained since 2013.

Two surfaces over one dataset:

Both are backed by the same services and return identical payloads — the only difference is the protocol envelope.

Search, ranking lists and site statistics need no credentials at all. Point a client at the URL and the first call works.

curl "https://www.myhuiban.com/api/conferences?ccf_rank=A&field=ai&submission_date_start=2026-08-05"

Add it to your MCP client

claude mcp add --transport http conference-partner https://www.myhuiban.com/mcp

https://www.myhuiban.com/mcp

{
  "mcpServers": {
    "conference-partner": {
      "url": "https://www.myhuiban.com/mcp"
    }
  }
}

{
  "servers": {
    "conference-partner": {
      "type": "http",
      "url": "https://www.myhuiban.com/mcp"
    }
  }
}

Ready-to-copy files for each client live in examples/mcp-config/, and the per-client walkthrough — including how to attach a key — is in docs/mcp-clients.md.

Also listed on Smithery, Glama and the official MCP registry as com.myhuiban.www/conference-partner.

Tools

Eight read-only tools, each with a REST twin returning the same payload.

Search accepts any combination of keyword, ccf_rank (A/B/C), core_rank (A*/A/B/C), qualis_rank, field (research areas, CS and beyond), submission/conference date ranges and updated_since for incremental sync. Every parameter is optional — filter-only queries such as "AI conferences with a deadline in the next 90 days" are the point.

Ranking lists cover open CFPs, the three tier lists, popularity, lowest acceptance rate, trending, upcoming/past — and for journals, special-issue CFPs and highest impact factor.

Prompts

The MCP server also ships prompts — ready-made tasks your client offers as one-click starting points, so you are not left staring at a list of tool names:

Full machine-readable definitions, refreshed from production by CI: spec/openapi.json and spec/mcp-tools.json.

Credentials

Create a key at myhuiban.com/account/api-keys — it never expires and can be revoked individually, which is what you want for a client that only lets you pin a fixed request header. A JWT from POST /api/login works too and draws on the same pool.

Every response carries X-Quota-Limit and X-Quota-Remaining. CORS is open, so browser JavaScript works.

401 = this endpoint needs a credential, or the one you sent is invalid. An unusable credential is never silently downgraded to the anonymous tier. 402 = account quota exhausted. 429 = anonymous per-IP quota exhausted.

Examples

Runnable, no API key required:

# upcoming CCF-A deadlines in AI, next 90 days
python examples/python/upcoming_deadlines.py --field ai --ccf A --days 90

# same thing in Node
node examples/node/upcoming-deadlines.mjs --field ai --ccf A --days 90

# build your own deadline snapshot (JSON + CSV + Markdown table)
python examples/python/export_deadlines.py --out ./snapshot

# incremental sync — only what changed since your last checkpoint
python examples/python/incremental_sync.py --since 2026-07-01

# shell one-liners
bash examples/curl/quickstart.sh

See examples/README.md.

Watch deadlines from your own repo

actions/deadline-watch is a GitHub Action that keeps one issue up to date with the deadlines you can still submit to — no API key, nothing to host. Issue #1 is this repo running it on a schedule, so you can see the output before adopting it.

- uses: sundou82/conference-partner-mcp/actions/deadline-watch@main
  with:
    field: security
    days: 120

Download the data

data/ carries a snapshot refreshed daily by CI, as JSON and CSV:

data/README.md renders the next 90 days as a table you can read in the browser.

The snapshot covers exactly what the anonymous tier already serves — nothing in it is gated. Per-venue detail (CFP full text, acceptance-rate history, edition history, ratings) stays behind the API, where a free key reaches it.

A file is stale the moment a deadline is extended, which in submission season is daily. Use it to browse, diff or bulk-load; call the API when it has to be right.

Subscribe to deadline changes, for free

CI commits data/ once a day and only when something actually changed, so the commit history of that directory is a change feed. GitHub publishes it as Atom, no account needed:

https://github.com/sundou82/conference-partner-mcp/commits/main/data.atom

Point a feed reader at it, or diff two commits to see exactly which deadlines moved:

git log --oneline -- data/
git diff HEAD~1 HEAD -- data/upcoming-deadlines.csv

Watching the repo on GitHub (Watch → Custom → Releases, or all activity) gets you the same signal as email.

Why this dataset

Deadline lists are easy to find. These are the parts that are hard to reproduce:

• Three ranking systems in one row — CCF, CORE and QUALIS on the same venue, so "CCF-A that is also CORE-A*" is one query rather than a join across three sources.
• A score for the venues those three leave blank — CCF, CORE and QUALIS between them cover about a fifth of the catalogue; for everything else the rank columns are empty. cp_index is our own 0–100 index over academic recognition, selectivity, longevity, community and how much a venue publishes about itself, so the long tail is comparable too. Every value ships with a confidence and a link to the method — a low score on a thinly documented venue means we know little about it, not that it is weak.
• Acceptance-rate history and edition history — how a venue's selectivity and location moved across editions, back to 2013.
• Journals as first-class entities — impact factor, publisher, ISSN, and special-issue CFPs, which conference-only lists do not carry.
• Incremental sync — updated_since with an updated_at watermark on every row, so a mirror is a diff rather than a re-crawl.
• Human-curated, not scraped: dates are entered and corrected by maintainers and by the organisers themselves through a venue-claim process.

Methodology, coverage and known limits: myhuiban.com/methodology.

About this repository

This repo is the distribution entry point for the API and MCP server: client configuration, examples, and CI-synced specs. The server itself is not open source — it is a hosted service; nothing here needs to be installed or run to use it.

• docs/mcp-clients.md — per-client setup, with and without a key
• docs/rest-api.md — endpoints, parameters, errors, incremental sync
• docs/data.md — what is in the dataset, attribution, terms
• examples/ — runnable Python, Node and curl clients; MCP config files
• spec/ — OpenAPI 3 document, MCP tool definitions and server card, re-fetched from production daily by CI so they cannot go stale

Authoritative reference: myhuiban.com/developers. Where this repo and that page ever disagree, the page is right — it ships with the API.

Issues and PRs about the client-side material here are welcome. For data corrections use the "report" link on the venue page, and for anything else email admin@myhuiban.com.

License

MIT for the code and configuration in this repository.

The data served by the API is not covered by that license. It is free to use, including commercially, with attribution: cite the canonical venue URL and "Conference Partner (myhuiban.com)". If your claim is about a CCF, CORE or QUALIS value itself, cite that body — Conference Partner reproduces those lists, it does not define them. cp_index is the exception: it is ours, so cite myhuiban.com/ranking and the algorithm version the row carries. Bulk use should start from data/ rather than a crawl of your own, and dataset licensing or model training needs a conversation first. See docs/data.md and the terms on the site.