ietf-llm 0.7.2

Maintain a local, LLM-queryable corpus of an IETF Working Group's public record (drafts, mailing list, GitHub issues, meetings), with an MCP server, semantic search, and NotebookLM export.

ietf-llm

Maintain a local, queryable corpus of an IETF Working Group's public record — charter, drafts, RFCs, meeting agendas, minutes, slides, transcripts, mailing list archives, and GitHub issues — for use with LLM-based tools.

What it's for

A working group's history is spread across mailing list archives, Datatracker, GitHub, and meeting materials — too much to hold in your head, and too scattered to search well by hand. With the record gathered into one queryable corpus, an LLM can help you:

• Get up to date with the state of discussions — what's open, what was recently decided, where a debate currently stands.
• Summarise the arguments already made about an issue — every distinct position on a topic, who holds it, and how the chairs ruled.
• Formulate a new proposal — surface the objections raised against similar ideas before, so you can anticipate them.
• Fact-check assertions about what's happened so far — grounded in the actual list traffic and chair statements, not someone's recollection.

Three supported workflows:

1. Use it as an MCP server — register ietf-llm-mcp with Claude, Codex, Gemini, Cursor, Zed, etc. and ask questions across any WG you've gathered.
2. Use it with NotebookLM — export the gathered corpus as a directory of clean text files (or push directly to NotebookLM Enterprise) and ingest it as a notebook source set.
3. Use it from the CLI — run semantic search over the cache directly with ietf-llm-search, no LLM client required.

Also works with IRTF Research Groups. Pass the RG's shortname (e.g. cfrg, hrpc, pearg) anywhere this README says <wg>.

Table of contents

• What it's for
• Installation
  • Shell completion
• Gather a corpus
• 1. Use as an MCP server
• 2. Use with NotebookLM
  • Export to a local directory
  • Export to NotebookLM Enterprise
• 3. Use from the CLI
• Reference
  • Commands
  • Gather options

Installation

pipx install ietf-llm

Behind a corporate firewall with TLS interception? Install with the certs extra:

pipx install ietf-llm[certs]

Shell completion

Optional. Add the line for your shell to its rc file to tab-complete commands, flags, and cached WG names:

# bash — in ~/.bashrc
eval "$(ietf-llm --completion bash)"

# zsh — in ~/.zshrc
eval "$(ietf-llm --completion zsh)"

# fish — in ~/.config/fish/config.fish
ietf-llm --completion fish | source

---

Gather a corpus

Each use mode below reads from a local cache. Gather it once per corpus; settings persist, so refreshing is a bare re-run (ietf-llm httpbis), and the semantic index updates incrementally each time.

ietf-llm httpbis --github httpwg/http-core --github httpwg/http-extensions

Everything lands in ~/.cache/ietf-llm/<name>/ — the single source of truth that the MCP server, NotebookLM exporter, and CLI search all read from.

First run: gathering a corpus also builds a local semantic-search index — this powers both the MCP search_corpus tool and ietf-llm-search. It uses BAAI/bge-small-en-v1.5 by default — ~130 MB, ~33M params, MPS-accelerated, runs entirely on your machine. Downloaded from Hugging Face on first use and cached; subsequent gathers reuse it. Override with --embed-model <id> for any model the llm package recognises. Pass --no-embed to skip the index (and the download) — useful for NotebookLM export or offline gathers.

A corpus doesn't have to be a Working Group — the name is classified automatically:

Command	Corpus
ietf-llm httpbis	a WG / RG / editorial WG / BoF: charter, drafts, meetings, ballots, list
ietf-llm last-call	a standalone mailing list (any archived at mailarchive.ietf.org — IETF, IRTF, or RFC-Editor)
ietf-llm rfced --mailing-list [e-mail]	a named list corpus (the address domain is optional)
ietf-llm new-ids --new-drafts --months 1	new Internet-Drafts in a rolling window
ietf-llm mnot --author [e-mail]	every draft a person has authored

See Gather options for the full flag set.

---

1. Use as an MCP server

ietf-llm-mcp is a stdio Model Context Protocol server that exposes the local corpus to any MCP-capable agent. Set up once, gather each WG you care about once, then ask questions indefinitely.

Register the server

Pick your client. The snippets below are correct as of writing — if your client has changed since, its own MCP docs are authoritative.

Gotcha (all clients): if ietf-llm-mcp was installed via pipx, the binary is on your shell PATH but may not be on the PATH inherited by a GUI app launched from Finder / Spotlight / Explorer. Use the absolute path (which ietf-llm-mcp) if the client can't find the command.

Claude Code

claude mcp add ietf-llm -- ietf-llm-mcp

Also install the bundled skill so Claude knows how to drive the tools well (digests before raw reads, search before slurping mailing-list files, etc.):

ietf-llm --install-claude-skill

Re-run after upgrading the package to pick up improvements.

Claude Desktop

Edit claude_desktop_config.json (create it if missing):

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json
• Linux: ~/.config/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "ietf-llm": {
      "command": "ietf-llm-mcp"
    }
  }
}

Quit and relaunch Claude Desktop — the config is only read at startup.

Codex CLI (OpenAI)

~/.codex/config.toml:

[mcp_servers.ietf-llm]
command = "ietf-llm-mcp"

Gemini CLI

~/.gemini/settings.json:

{
  "mcpServers": {
    "ietf-llm": {
      "command": "ietf-llm-mcp"
    }
  }
}

opencode

~/.config/opencode/opencode.json (or opencode.json in your project root):

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "ietf-llm": {
      "type": "local",
      "command": ["ietf-llm-mcp"],
      "enabled": true
    }
  }
}

Cursor

In-app MCP settings panel, or ~/.cursor/mcp.json (global) or .cursor/mcp.json (per-project):

{
  "mcpServers": {
    "ietf-llm": {
      "command": "ietf-llm-mcp"
    }
  }
}

Zed

~/.config/zed/settings.json:

{
  "context_servers": {
    "ietf-llm": {
      "command": {
        "path": "ietf-llm-mcp",
        "args": []
      },
      "settings": {}
    }
  }
}

Tuning. Each tool call has a server-side deadline so a stuck call fails fast with a clear message rather than hanging to the client's timeout. It defaults to 120 seconds; override (or disable, with 0) by setting IETF_LLM_TOOL_TIMEOUT in the server's environment — e.g. add "env": {"IETF_LLM_TOOL_TIMEOUT": "180"} to the JSON config above.

---

2. Use with NotebookLM

NotebookLM ingests a corpus as a set of source files. ietf-llm-export turns the gathered cache into an upload-ready directory, or pushes straight to a NotebookLM Enterprise notebook.

Workflow note: export always produces a complete fresh dump. Create a new notebook on each refresh rather than trying to merge updates into an existing one.

Pass --no-embed when gathering if you only plan to use NotebookLM — it has its own index, so the local one is wasted work:

ietf-llm httpbis --no-embed \
    --github httpwg/http-core --github httpwg/http-extensions

Export to a local directory

ietf-llm-export httpbis --destination ~/notebooklm/httpbis

Drag the directory's contents into NotebookLM as sources. Per-thread mailing list conversations and per-issue GitHub records are bundled by year / repo to stay under NotebookLM's 50-source free / 300-source Plus limit.

Export to NotebookLM Enterprise

If you have Google Workspace Enterprise with NotebookLM enabled, ietf-llm-export can create the notebook and upload sources directly:

ietf-llm-export httpbis --create my-gcp-project-id

One-time setup:

1. Google Cloud Project with the Discovery Engine API enabled.
2. OAuth credentials: create an "OAuth 2.0 Client ID" (Desktop App) in the Cloud Console.
3. Save the JSON as client_secrets.json in ~/.config/ietf-llm/ (or pass --credentials-file PATH).

First run opens a browser to authorise; the token is cached at ~/.config/ietf-llm/token.json.

Per-WG export settings are persisted at ~/.config/ietf-llm/<wg>/export.json — subsequent runs of the same mode need only ietf-llm-export <wg>.

---

3. Use from the CLI

ietf-llm-search runs semantic search over a gathered corpus and prints ranked excerpts to stdout — useful on its own, or as input to another tool.

ietf-llm-search httpbis "skepticism about cookie partitioning" -k 8

Chunks are content-aware: one chunk per mailing list message, one per issue comment, and a windowed slice of drafts/RFCs/transcripts. The index lives at ~/.cache/ietf-llm/<wg>/embeddings.db and updates incrementally on each gather — see Gather a corpus for the embedding model.

---

Reference

Commands

Command	Job	Reads	Writes
ietf-llm	Gather / refresh a corpus	network	cache
ietf-llm-export	Mirror cache to dir, or push to NotebookLM Enterprise	cache	dir / NotebookLM
ietf-llm-search	Semantic search over the cache	cache	stdout
ietf-llm-mcp	Expose the cache to MCP clients	cache	stdio (MCP)

All four are independent. The cache (~/.cache/ietf-llm/<wg>/) is the single source of truth; everything else reads from it.

Gather options

ietf-llm [OPTIONS] <name>

<name> is the corpus to gather, classified automatically:

• a Working Group / Research Group / editorial WG / BoF shortname (httpbis, cfrg, rswg) — gathered in full (charter, drafts, meetings, ballots, mailing list);
• a mailing list archived at mailarchive.ietf.org — IETF, IRTF, or RFC-Editor (last-call, irtf-discuss, rfc-interest) — that list on its own;
• any other label given explicit sources (--draft / --mailing-list / --github / --new-drafts / --author);
• prefix with x- to skip the Datatracker group lookup entirely (a fully manual corpus).

A name that is none of these and has no configured sources is rejected as a likely typo.

Sources (what to gather; all repeatable / persisted):

• --github OWNER/REPO — a GitHub repo whose issues to include.
• --draft DRAFT-NAME — an extra Internet-Draft to track, beyond a WG's own documents. Version suffix stripped; every revision gathered.
• --mailing-list LIST — an extra list to sync (any archived at mailarchive.ietf.org). A bare name or a full address; the domain is optional and ignored (rswg, [e-mail]).
• --new-drafts — subscribe to new Internet-Drafts: every -00 submitted within --months (rolling window; drafts age out).
• --author PERSON — every draft PERSON authored. PERSON is an email ([e-mail], recommended), a Datatracker person id, or an exact full name. Drafts only.
• --add-mentioned-drafts — also pull drafts the corpus's threads/issues mention but don't already include. Sticky.

Scope & filtering:

• --months N — months of mailing list / meeting / new-draft history (default 12).
• --github-label LABEL / --exclude-github-label LABEL — include / exclude issues by label.

Digests & search index:

• --summarize / --summarize-model MODEL — add LLM-generated one-liners to digests via the llm package.
• --no-embed — skip the semantic search index (it backs ietf-llm-search and the MCP search_corpus tool). On by default, incremental.
• --embed-model MODEL — embedding model id (default: a small local model).
• --rebuild-embeddings — drop and re-embed everything instead of the incremental update.

Cache & config:

• --list — list cached corpora (name, kind, status, last-gathered, and a one-line subject — the group name, list, or tracked author), then exit.
• --clear-cache — wipe this corpus's cache and re-download.
• --clear-config — clear this corpus's persisted config.
• --quiet / --verbose.

Per-corpus settings are persisted at ~/.config/ietf-llm/<name>/gather.json.

GitHub auth. Set GITHUB_TOKEN on the gather invocation (a fine- scoped read-only token is plenty); without one you'll hit anonymous API rate limits quickly on large WGs. Prefer inline-passing over exporting in your shell rc so the token doesn't leak into every other subprocess:

GITHUB_TOKEN=ghp_... ietf-llm httpbis
# or, from a secret manager:
GITHUB_TOKEN=$(security find-generic-password -s github-readonly -w) \
    ietf-llm httpbis