California Jury Instructions MCP Server

@pipeworx/california-jury-instructions

Official California jury instructions as adopted by the Judicial Council of California — CACI (civil, e.g. "303" = Breach of Contract—Essential Factual Elements) and CALCRIM (criminal, e.g. "200" = Duties of Judge and Jury), current (2026) edition, full instruction text with Directions for Use/Bench Notes, Sources and Authority/Authority, and revision history.

Part of Pipeworx — an MCP gateway connecting AI agents to 1715+ live data sources. This is an independent, unofficial integration — not affiliated with, endorsed by, or published by the upstream provider.

Tools

• caci_lookup(number) — full official text of one CACI (civil) instruction or verdict form by number ("303", "VF-303"): instruction text, Directions for Use, Sources and Authority, Secondary Sources, revision history.
• calcrim_lookup(number) — full official text of one CALCRIM (criminal) instruction or verdict form by number ("200", "VF-200"): instruction text, Bench Notes (incl. Instructional Duty), Authority, Related Issues, Secondary Sources, Commentary where present, revision history.
• ca_ji_search(query, set?) — find the official instruction NUMBER for a claim or charge by keyword ("breach of contract" -> CACI No. 303), across CACI, CALCRIM, or both. Title search, not full-text search.
• ca_ji_editions(set?) — current edition year/adoption date and archived edition PDF links for either or both sets.

Auth

Keyless. This reads the Judicial Council's own published, public PDFs — not a third-party API, no registration of any kind.

Data sources

• https://www.courts.ca.gov/system/files/file/judicial_council_of_california_civil_jury_instructions_2026.pdf — CACI (civil), current 2026 edition, the single source for every caci_lookup response.
• https://courts.ca.gov/system/files/file/calcrim-2026.pdf — CALCRIM (criminal), current 2026 edition, the single source for every calcrim_lookup response.
• https://courts.ca.gov/partners/juryinstructions.htm — the Judicial Council's own index page linking to both resource centers, current editions, and archived editions back to ~2019-2021 (surfaced via ca_ji_editions, not text-extracted).

Why this pack bakes the index instead of fetching per call

Both editions are published as ONE full-edition PDF — CACI 2026 is ~3,560 pages/11MB, CALCRIM 2026 is ~2,680 pages/84MB. There is no per-instruction endpoint, and parsing either document takes tens of seconds even outside a Worker, so scripts/bake-index.mjs downloads and parses both ONCE (re-run when a new edition is adopted, roughly annually) into src/ca-ji-index-data.ts, which the gateway uploads to KV and injects as args._bakedIndex on first call per isolate — see that script's header and src/index.ts's top comment for the full mechanism (same pattern as south-dakota-code, hawaii-code).

shared/src/pdf-text.ts does not work on these PDFs. Tested directly against the real CACI PDF: it silently drops every inter-word space (this typesetter encodes word gaps as TJ-array kerning adjustments, not space glyphs — e.g. "Thiseditionof CACI includes...") and mis-decodes several ligatures (fi/fl -> "®", en-dash -> "±") via subsetted font encodings outside its documented scope. pdftotext (poppler-utils) reconstructs spacing correctly via real glyph positioning and is used instead — a build-time dependency only; the Worker never runs the bake script. Re-running it (to pick up a new edition) needs pdftotext on the machine that runs it (brew install poppler / apt install poppler-utils).

Parsing approach: both documents' running headers name the instruction on every content page ("CACI No. 303", "CALCRIM No. 200"), and every series/category has its own divider page listing every instruction number + official title — used as the ground-truth index, independent of body-text parsing. Verified 2026-10-07 against the live 2026 editions: CACI 1,140/1,140 ground-truth numbers accounted for (1 — #4001 — correctly revoked, not a parse gap); CALCRIM 780/780, 100%.

Known limits (said plainly, not papered over):

• Only the CURRENT edition's text is baked. Archived editions (back to ~2019-2021) are linked by URL via ca_ji_editions, not text-extracted — a past edition's exact wording is not retrievable from this pack.
• "Proposed revisions" (the Council's separate public-comment process for pending changes) are NOT covered at all. Everything served is ADOPTED, in-force text (status: "adopted" on every successful lookup).
• ca_ji_search matches official instruction/verdict-form TITLES and series names, not full instruction body text.

Licensing: both documents carry an explicit reprint grant on their own copyright page — "In accord with Cal. Rules of Court, Rule 2.1050(c), these instructions and verdict forms are freely available for use and reproduction, with appropriate citation to this publication." This pack cites the publication via source/source_url on every response.

Quick Start

Add to your MCP client (Claude Desktop, Cursor, Windsurf, etc.):

{
  "mcpServers": {
    "california-jury-instructions": {
      "url": "https://gateway.pipeworx.io/california-jury-instructions/mcp"
    }
  }
}

What this endpoint actually serves

tools/list at https://gateway.pipeworx.io/california-jury-instructions/mcp returns the tools in the table above plus the shared Pipeworx meta-tools — ask_pipeworx, discover_tools, search_within, remember/recall and the rest of the gateway-wide set. So the tool count you see is larger than this table: a single-pack endpoint currently lists roughly 30 shared tools alongside the pack's own. The connection's initialize response states its exact scope, and is the authoritative answer for a given day.

This is deliberate, not multiplexing by accident. The meta-tools are what let a scoped connection answer a question this pack does not cover — via ask_pipeworx, which routes across the whole catalog — without you adding a second MCP server. There is currently no way to mount a pack endpoint without them; if the extra schemas cost you more context than the routing is worth, connect to the full gateway once rather than to several pack endpoints.

Or connect to the full Pipeworx gateway to get every pack's tools listed directly, instead of just this one's:

{
  "mcpServers": {
    "pipeworx": {
      "url": "https://gateway.pipeworx.io/mcp"
    }
  }
}

Both URLs reach the same gateway and the same 1715+ data sources. The only difference is which pack's tools are listed directly; ask_pipeworx reaches all of them from either one.

No MCP client? Call it over HTTP

curl -X POST https://gateway.pipeworx.io/v1/tools/caci_lookup \
  -H 'Content-Type: application/json' \
  -d '{"number":"303"}'

No account needed for the first calls. Inspect any tool: GET https://gateway.pipeworx.io/v1/tools/caci_lookup. Find one: POST https://gateway.pipeworx.io/v1/tools/search_packs with {"query":"..."}.

Standalone (no gateway account)

This package also runs as a local stdio MCP server — no Pipeworx account, no gateway round-trip:

{
  "mcpServers": {
    "california-jury-instructions": {
      "command": "npx",
      "args": ["-y", "@pipeworx/mcp-california-jury-instructions"]
    }
  }
}

Or run it directly to confirm it starts:

npx -y @pipeworx/mcp-california-jury-instructions

It speaks MCP over stdin/stdout and answers initialize/tools/list/tools/call for only this pack's tools — none of the shared meta-tools the gateway connection above adds. Same source, same tools, no ask_pipeworx routing.

Using with ask_pipeworx

Instead of calling tools directly, you can ask questions in plain English — this works on the pack endpoint above as well as on the full gateway:

ask_pipeworx({ question: "your question about California Jury Instructions data" })

The gateway picks the right tool and fills the arguments automatically.

More

• Docs and guides
• pipeworx.io

License

MIT