Repository avatar
Search & Data Extraction
v0.2.0
active

oeis-mcp-server

io.github.cyanheads/oeis-mcp-server

Identify integer sequences by terms, search the OEIS, read formulas, programs, b-files, cross-refs.

Documentation

@cyanheads/oeis-mcp-server

Identify integer sequences by terms, search the OEIS, read formulas, programs, b-files, cross-refs via MCP. STDIO or Streamable HTTP.

6 Tools • 1 Sequence Resource • 1 Skill

Version License Docker MCP SDK npm TypeScript Bun

Install in Claude Desktop Install in Cursor Install in VS Code

Framework

Public Hosted Server: https://oeis.caseyjhand.com/mcp


Overview

Integer sequences from the On-Line Encyclopedia of Integer Sequences (OEIS). Identify a sequence from a few observed terms, search with OEIS's own query syntax, read an entry's formulas and programs, page through extended terms from its b-file, and walk its cross-references. Every result links to its oeis.org page. No API key required.

Tools

ToolDescription
oeis_identify_sequenceIdentify sequences that contain a run of consecutive terms, in OEIS relevance order, with the index where the run begins
oeis_search_sequencesSearch with OEIS query syntax: words, quoted phrases, term lists, and prefixes such as keyword: and author:
oeis_get_sequenceFetch one entry by A-number: terms, offset, keywords, formulas, programs, comments, references, links, and cross-references
oeis_get_termsList terms a(n) with their indices from the entry's b-file, or from its data line when it has none
oeis_get_cross_refsList the sequences an entry cross-references, or the entries that mention it
oeis_list_referenceDecode keyword flags, search syntax, and identifier and offset conventions, with no upstream call

Resources

ResourceDescription
oeis://sequence/{aNumber}One entry by A-number as JSON, always whole

The same record is available through oeis_get_sequence for clients that only call tools.

Skill

The OEIS guide covers candidate verification, formula matching, and partial-result coverage. Plugin installs include it; MCP clients can read skill://oeis-guide/SKILL.md.

Capability reference

oeis_identify_sequence tool

  • terms: up to 60 consecutive terms separated by commas or spaces (a bracketed list or a trailing ... is accepted), each an integer of at most 200 digits or _ for one unknown term; matchSigns (default false) ignores signs unless set
  • Up to 10 candidates per page, start 0–100 in steps of 10; each row carries the oeis_search_sequences row fields plus matchStartIndex, the n where the run begins in that entry's data line (absent when the run isn't there)
  • No match and "too many results" both return an empty page; the notice says which and how to retry (drop leading terms, divide out a common factor, or add terms). OEIS matches a run only within each entry's data line (at most about 270 characters), never its b-file, and a zero-hit notice says so when every nonzero term has 10 or more digits, five or more terms of 4 or more digits lie within 10% of each other, or the run passes 270 characters

oeis_search_sequences tool

  • query (1–1,000 characters) is sent as written: words, "quoted phrases", term lists, prefixes such as keyword:, author:, name:, formula:, xref:, and id:, | for OR, and a leading - to exclude
  • sort: relevance (default), number, created, or modified; 10 per page, start 0–100 in steps of 10
  • Rows carry name, terms, offset / firstIndex, keywords, author, legacyIds, modified (the last edit, ISO 8601 with offset), and url, in oeis_get_sequence's formats; author and legacyIds are absent where OEIS has none
  • Reports totalCount and effectiveQuery, the query as OEIS parsed it. A query OEIS parses as one run of terms (effectiveQuery reads seq: or signed: and the terms, as for bare numbers separated by commas) is matched within each entry's data line, as in oeis_identify_sequence, and its zero-hit notice names that limit under the same conditions
  • With spaces alone between numbers, OEIS starts a new run at a negative term after the first and does not join fewer than three numbers; when effectiveQuery shows that split, the notice says so and gives the terms joined by commas to search as one run

oeis_get_sequence tool

  • Returns name, terms, offset / firstIndex, keywords, author, legacyIds, referenceCount, revision, created / modified, url, and bFileUrl, plus eight sections: comments, formulas, examples, programs, references, links, crossReferences, extensions
  • When the sections exceed 24,000 characters of serialized JSON, kind: "outline" returns the core fields and a sized sections list; pass sections (e.g. ["formulas", "programs"]) to get the chosen ones
  • A selection past 100,000 bytes comes back in parts cut between whole items; pass each part's nextFromItem as fromItem, with the same sections, for the next. A fromItem naming a section not in sections fails as from_item_not_selected
  • An unknown A-number fails as sequence_not_found; withdrawn (dead), reserved, and recycled entries come back with a notice

oeis_get_terms tool

  • fromIndex defaults to the first available index. limit defaults to 100 and applies up to 1,000 terms per call, regardless of digit count. Pass nextFromIndex to continue.
  • source: "bfile" reads the entry’s b-file in 1 MiB parts. The tool can seek up to two parts beyond the first and read one adjacent part to fill the slice. It joins adjacent terms only when their indices continue and the ETag matches. If the optional part is unavailable, the tool returns the terms it has with a continuation cursor. bFileCut: true means the file may contain terms beyond lastAvailableIndex. source: "data" means the entry has no b-file and the terms are its data line.
  • Terms come back as { n, value }, value an exact decimal string; an unknown A-number fails as sequence_not_found
  • A b-file line that is not blank, a # comment, or an n a(n) pair is skipped, and the notice says how many; a b-file that holds no pair at all fails as upstream_unparseable and is not cached
  • If a line at a part boundary exceeds the 4 KiB overlap and cannot be read whole, the notice points to the b-file.

oeis_get_cross_refs tool

  • direction: "outgoing" (default) lists the A-numbers named in the entry's cross-reference lines, with the note written beside each and the lines verbatim; "incoming" lists the entries that mention the A-number
  • 10 rows per page, start 0–100 in steps of 10; every row carries url, and resolved rows add the name, terms, offset, keywords, author, legacy IDs, and last edit (modified)
  • If the name lookup for an outgoing page is rate-limited, unavailable, or times out, rows come back resolved: false with a retry notice instead of failing; sequence_not_found applies to outgoing only

oeis_list_reference tool

  • topic: keywords (flags such as nonn, core, tabl, cons), search_syntax (prefixes, operators, wildcards, sort orders, paging), or identifiers (A-numbers, legacy M/N numbers, offsets, b-files)
  • Returns entries (name, description) and notes from static tables

oeis://sequence/{aNumber} resource

  • The full entry as application/json: the record oeis_get_sequence returns with kind: "full", never outlined
  • aNumber takes the same forms as the tools (oeis://sequence/A108 reads A000108); an unknown A-number fails as sequence_not_found

Features

Built on @cyanheads/mcp-ts-core: stdio and Streamable HTTP transports, pluggable auth (none / jwt / oauth), swappable storage (in-memory, filesystem, Supabase, Cloudflare KV/R2/D1), structured logging with optional OpenTelemetry tracing.

OEIS-specific:

  • One paced, cached oeis.org client (see Pacing and caching); searches tell "no results" apart from "too many results" and report the match total
  • Forgiving A-numbers: A000045, a000045, A45, 45, and an oeis.org sequence URL all resolve in every tool; a legacy M/N book number is pointed to oeis_search_sequences
  • Exact terms: every term and b-file value is a decimal string, so values past 2^53 keep every digit
  • Pasted term runs are normalized: brackets, a trailing ..., the Unicode minus sign, and mixed comma and space separators

Agent-friendly output:

  • Paging context on every page (truncated, shown, cap, plus totalCount and effectiveQuery where OEIS reports them) and a notice that says how to narrow or continue
  • Typed failures with recovery hints: sequence_not_found, from_item_not_selected, upstream_unparseable, pacer_shed, and upstream_rate_limited, the rate-limit errors carrying retryAfter when one is known
  • Discriminated outputs (kind, source, per-row resolved), so callers branch on data, not string parsing
  • Every record and row carries its https://oeis.org/A###### URL; contributor-written text renders as code spans and code fences, so a markdown client shows formulas and names exactly as written, and only http and https link URLs are kept

Pacing and caching

oeis.org asks automated clients to wait 10 seconds between requests. The server sends one upstream request at a time and starts each at least 10 seconds after the last, across every tool. Results are cached in memory, so a repeated lookup returns at once and costs oeis.org nothing: records for 24 hours, or until a search result shows a later edit (then revalidated with a conditional request), search pages for 1 hour, and b-file parts for 7 days. An entry's cached "no b-file" answer is dropped as soon as a fetched record lists a b-file.

A call with nothing cached waits its turn in the queue. When the wait would pass OEIS_QUEUE_MAX_WAIT_MS (default 30,000 ms), the call fails with pacer_shed and a retryAfter instead of hanging. If oeis.org answers 429, the server holds every queued call back before trying again, and reports upstream_rate_limited once its retries are spent. The pace is per server process.

Known limitations

  • Anonymous paging is capped upstream at 110 results per query (start ≤ 100); A000045's 6,161 incoming references, for example, are reachable only through their first 110.
  • A query OEIS judges too broad returns no rows ("Too many results"), which is common for a single word (prime) or a two-term run. The server reports it but cannot page it.
  • OEIS searches an unknown prefix as plain words instead of rejecting it; the zero-hit notice names it.
  • Term matching (oeis_identify_sequence, and an oeis_search_sequences query of bare numbers or seq: / signed: terms) searches each entry's data line only, not b-files, and data lines are short: at most 269 characters, signs included, across 105 measured (2026-10-01). A run that starts beyond the data line, or one longer than about 270 characters, is not found.
  • oeis_get_terms shares a 50 s deadline across its reads. If two seek reads do not reach the requested index, call again to continue from cached parts. Failures on required reads and caller cancellation propagate. A b-file served without byte ranges or a strong ETag is read only to its first 1 MiB.
  • oeis_get_cross_refs outgoing shares the start ≤ 100 bound, so an entry naming more than 110 distinct A-numbers lists only the first 110; lines still names them all.

Data and licensing

OEIS content is licensed CC BY-SA 4.0 under the OEIS End-User License Agreement. When you reuse it, credit "The On-Line Encyclopedia of Integer Sequences" with a link to https://oeis.org/ or to the sequence page; every record the server returns carries its https://oeis.org/A###### URL for that purpose. ShareAlike applies: adaptations of OEIS content must be shared under the same license.

This server is an independent project and is not affiliated with or endorsed by the OEIS Foundation.

Getting started

Public Hosted Instance

A public instance is available at https://oeis.caseyjhand.com/mcp — no installation required. Point any MCP client at it via Streamable HTTP:

{
  "mcpServers": {
    "oeis-mcp-server": {
      "type": "streamable-http",
      "url": "https://oeis.caseyjhand.com/mcp"
    }
  }
}

Every caller of the hosted instance shares one oeis.org pace of one request every 10 seconds; cached lookups return at once. For sustained use, run your own instance.

Self-Hosted / Local

Add the following to your MCP client configuration file.

{
  "mcpServers": {
    "oeis-mcp-server": {
      "type": "stdio",
      "command": "bunx",
      "args": ["@cyanheads/oeis-mcp-server@latest"],
      "env": {
        "MCP_TRANSPORT_TYPE": "stdio",
        "MCP_LOG_LEVEL": "info"
      }
    }
  }
}

Or with npx (no Bun required):

{
  "mcpServers": {
    "oeis-mcp-server": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@cyanheads/oeis-mcp-server@latest"],
      "env": {
        "MCP_TRANSPORT_TYPE": "stdio",
        "MCP_LOG_LEVEL": "info"
      }
    }
  }
}

Or with Docker:

{
  "mcpServers": {
    "oeis-mcp-server": {
      "type": "stdio",
      "command": "docker",
      "args": ["run", "-i", "--rm", "-e", "MCP_TRANSPORT_TYPE=stdio", "ghcr.io/cyanheads/oeis-mcp-server:latest"]
    }
  }
}

For Streamable HTTP, set the transport and start the server:

MCP_TRANSPORT_TYPE=http MCP_HTTP_PORT=3010 bun run start:http
# Server listens at http://localhost:3010/mcp

Prerequisites

  • Bun v1.4.4 or higher (or Node.js v24+).
  • No API key or account: the server reads public oeis.org data.

Installation

  1. Clone the repository:
git clone https://github.com/cyanheads/oeis-mcp-server.git
  1. Navigate into the directory:
cd oeis-mcp-server
  1. Install dependencies:
bun install
  1. Configure environment:
cp .env.example .env
# optional: adjust OEIS_QUEUE_MAX_WAIT_MS, the transport, or the log level

Configuration

VariableDescriptionDefault
OEIS_QUEUE_MAX_WAIT_MSLongest a call waits in the oeis.org request queue before failing with pacer_shed and a retryAfter, in ms. Each call's 50 s deadline also has to cover the request itself, so waits stop at about 35,000 ms whatever the setting.30000
MCP_TRANSPORT_TYPETransport: stdio or http.stdio
MCP_HTTP_PORTHTTP server port.3010
MCP_SESSION_MODEHTTP session mode: stateless, stateful, or auto.stateless
MCP_AUTH_MODEAuthentication: none, jwt, or oauth.none
MCP_JWT_EXPECTED_AUDIENCEAudience claim required in JWT mode. The server warns at startup when JWT mode is configured without an audience.unset
MCP_JWT_EXPECTED_ISSUERIssuer claim required in JWT mode. Set it alongside the expected audience.unset
MCP_LOG_LEVELLog level (debug, info, warning, error, etc.).info
LOGS_DIRDirectory for log files (Node.js only).<app-root>/logs
STORAGE_PROVIDER_TYPEStorage backend: in-memory, filesystem, supabase, cloudflare-kv/r2/d1.in-memory
OTEL_ENABLEDEnable OpenTelemetry.false

See .env.example for every server setting and the common framework overrides.

Running the server

Local development

  • Build and run the production version:

    # One-time build
    bun run rebuild
    
    # Run the built server
    bun run start:http
    # or
    bun run start:stdio
    
  • Run checks and tests:

    bun run devcheck  # Lints, formats, type-checks, and more
    bun run test      # Runs the test suite
    

Project structure

DirectoryPurpose
src/index.tscreateApp() entry point: registers the tools and resource, sets the server instructions, and starts the OEIS service.
src/configServer-specific environment variable parsing and validation with Zod.
src/mcp-server/toolsTool definitions (*.tool.ts). Six tools.
src/mcp-server/resourcesResource definitions. The oeis://sequence/{aNumber} resource.
src/mcp-server/sharedSchemas shared across definitions (A-number input, paging, summary rows), Markdown helpers for contributor-written text, and the data-line-limit test and sentence the term-matching zero-hit notices share.
src/services/oeisoeis.org client: pacer, retries, fetch boundaries, LRU cache, record normalization, and the internal-format parser.
skills/Published OEIS guide, included in npm, Docker, and MCPB and served over MCP.
tests/Unit tests mirroring the src/ structure; upstream responses come from fixtures, never the live site.

Development guide

See CLAUDE.md for development guidelines and architectural rules. The short version:

  • Handlers throw, framework catches — no try/catch in tool logic
  • Use ctx.log for logging and ctx.enrich for notices, totals, and paging context
  • Register new tools and resources in the barrels at src/mcp-server/*/definitions/index.ts
  • Wrap external API calls: validate raw → normalize to domain type → return output schema; never fabricate missing fields

Contributing

Issues are welcome. Run checks and tests before submitting:

bun run devcheck
bun run test

License

This project is licensed under the Apache 2.0 License. See the LICENSE file for details.