v0.6.3

Automated PATCH refresh — upstream tc39/* moved. Same code, new spec data. See spec.about for the current SHAs.

v0.6.2

Automated PATCH refresh — upstream tc39/* moved. Same code, new spec data. See spec.about for the current SHAs.

v0.6.1

Automated PATCH refresh — upstream tc39/* moved. Same code, new spec data. See spec.about for the current SHAs.

v0.6.0

Surfaces a clause's citations to external specs (Unicode, IETF, WHATWG)
as structured, resolvable URLs.

Added

• External-spec citations on clause.get. A new optional
  external_refs field on the returned Clause lists a clause's outward
  links to normative external specs — Unicode TRs, IETF RFCs, WHATWG — as
  { url, text }, filtered to a normative-host allowlist. Present only
  when the clause cites something external (e.g.
  sec-string.prototype.normalize → the Unicode Normalization Forms
  reference).
• external category on spec.crossrefs. The crossref graph now
  carries the same citations as an outgoing external category alongside
  the internal in/out hits, so tracing a clause's references surfaces both
  internal targets and outward spec links in one call. Present on
  direction: "out" / "both".

Changed

• Install + run docs are now agent-agnostic. The setup instructions
  present the two transports — stdio (npx) and hosted HTTP
  (mcp.xyzzylabs.ai/tc39) — for any MCP client, instead of leading with
  one specific client's config.

v0.5.0

Adds the MCP prompts capability and moves the hosted deployment to its
own domain.

Added

• MCP prompts (workflow templates). Seven reusable prompts —
  explain-clause, compare-editions, find-and-read,
  trace-crossrefs, proposal-status, test262-for-feature,
  cite-reproducibly — that steer an agent through the right tool
  sequence. Pure string templates (no execution); advertised on both the
  stdio server and the hosted Worker via prompts/list + prompts/get.

Changed

• Hosted deployment moved to mcp.xyzzylabs.ai/tc39. The docs site,
  the registry MCP endpoint, and the npm package's R2-fetch origin point
  there now. The old *.workers.dev URL still serves identical data.
• Edition aliases normalize. ES2025, 2025, es-2025, Latest,
  MAIN resolve to canonical editions instead of 404-ing on a missing
  snapshot key — the same tolerance the ecma262 / ecma402 spec
  aliases got in 0.4.1. A well-formed but out-of-range edition (es2015)
  still reports "unsupported".
• Discoverability: step-text + test262 linkage. The server
  instructions now point agents at spec.search { search_steps: true }
  for "where is X invoked" queries and at test262.search { esid } for a
  clause's conformance tests (a clause id is its esid), and drop the
  hard-coded tool list — the live set comes from tools/list.

v0.4.1

A bug-fix and polish release.

Fixed

• Spec aliases now resolve instead of 404-ing. Tools accept the
  long spec names agents commonly pass — ecma262 / ecma402 (plus
  ECMA-262, es, intl) — and normalize them to the canonical
  262 / 402 before building the snapshot key. Previously
  spec: "ecma262" produced a confusing Missing parsed spec object in R2: spec-ecma262-… error; an unrecognized spec now returns a
  clear Unknown spec … message.
• Clean stdio stream from the documented dev launch config. The
  .mcp.json wiring that runs the server from local source now passes
  npm run --silent mcp, so npm's lifecycle banner can't leak onto
  stdout and corrupt the JSON-RPC stream a stdio MCP client reads.

Changed

• Reworded the project disclaimer from "unofficial / community
  project" to "independent project — not an official Ecma International
  or TC39 publication," across the README, docs landing page,
  agent-facing instructions, and the server.json registry
  description.

v0.4.0

The hosted Cloudflare Worker grows from 6 to 17 of the 19 tools. Every
newly-hosted tool shares its logic with the stdio server through a
dependency-free src/spec/* (or src/index/*) module, so the two
transports answer identically and can't drift.

Added

• Eleven more tools on the hosted Worker, each reading the
  parsed-spec JSON or index it already loads from R2:

  • spec.grammar, spec.tables, spec.sdo_index — grammar
    productions from <emu-grammar>, <emu-table> content, and the
    Syntax-Directed-Operation-by-production index.
  • clause.outline, spec.global_search — the section tree, and one
    search across both ECMA-262 and ECMA-402.
  • spec.snapshots — the live (spec, edition, sha, fetched_at)
    snapshots the Worker serves from R2.
  • spec.symbol_resolve, spec.well_known_intrinsics — resolve
    notation ([[Slot]], %Intrinsic%, ~enum~) and enumerate
    well-known intrinsics with their defining clauses.
  • spec.diff — clause-level diff across two editions of a spec.
  • spec.crossrefs — incoming / outgoing references, with the
    AOID-densified reverse index and the opt-in 262 ↔ 402 cross-spec
    pass.
  • test262.search — ranked search over the tc39/test262 index,
    served from the same R2 side-index spec.about already reads.

  The two tools that stay stdio-only are spec.history (shells out to
  git log against a vendored checkout) and test262.get (reads each
  test's full source from the vendored test262 corpus, which isn't in
  R2).

Changed

• Every ported tool's logic now lives in a shared, dependency-free
  src/spec/* (or src/index/*) module imported by both the stdio
  server and the bundled Worker, replacing what would otherwise be a
  hand-maintained second copy. The stdio tool surface — every schema and
  result shape — is unchanged; this extends the 0.3.1 consolidation
  across every newly-hosted tool.
• The hosted-vs-stdio tool split is a single source of truth. Both
  transports' server instructions and the Worker's tools/list
  registry derive their tool lists and counts from one tool_inventory
  module, so a tool crossing over updates every surface at once.
• spec.about's metadata scan no longer evicts the Worker's hot
  parsed-spec cache. The introspection scan that reads every snapshot
  for its pin now uses a parse-and-discard path instead of thrashing
  the capacity-4 LRU that clause.get / spec.search depend on.
• The docs now mark transport availability. The tool reference
  carries a per-tool Availability line (hosted Worker vs
  stdio-only), generated from the same tool_inventory source of
  truth, and getting-started splits into self-contained Local (stdio)
  and Hosted (HTTP) walkthroughs.

v0.3.1

The hosted Cloudflare Worker reaches feature parity with the stdio
server, and the stdio ↔ Worker code is unified so the two can't drift
apart again.

Fixed

• spec.search on the hosted Worker now honors search_steps. It
  previously ranked only aoid / title / id matches; step-text matches
  (the steps tier) were silently dropped. The stdio server already
  did this — now both transports rank a query identically.
• proposal.list on the hosted Worker now accepts the spec filter
  (262 / 402). It was stdio-only, so the Worker couldn't narrow
  proposals to one spec.

Changed

• The spec/edition catalog, the spec.search ranking, and the
  proposal.list filter are now single shared modules used by both
  the stdio server and the Worker, replacing hand-maintained copies
  that had drifted. The stdio tool surface is unchanged
  (docs/tools.md is identical) — this is internal consolidation plus
  the two Worker fixes above.

v0.3.0

The es2026 edition lands on both specs.

Added

• ECMA-262 and ECMA-402 add the es2026 edition. Both specs now
  cover es2016 – es2026 plus main. tc39/ecma262 tags es2026;
  tc39/ecma402 publishes it as an esYYYY branch. clause.get,
  spec.search, spec.diff, spec.history, and the rest of the
  edition-aware surface resolve es2026 on both specs.

Changed

• latest now resolves to es2026 on both ECMA-262 and ECMA-402
  (previously es2025). main / draft / next continue to address
  the working draft.
• The npm bundle tracks the new stable. The offline fallback now
  ships spec-262-es2026 + spec-402-es2026 (replacing the es2025
  pair); every other edition is fetched from the hosted Worker on
  demand.

v0.2.0

The data model moves
from "ship every snapshot in the tarball" to
"cache on first use, fetch from the hosted Worker, fall back to a
bundled subset." The npm version stops tracking spec data and starts
meaning code again.

Added

• ECMA-402 reaches edition parity with ECMA-262. 402 publishes
  each annual edition as an esYYYY branch (not a tag); the catalog
  now exposes the full es2016 – es2025 range plus main, where it
  previously had only es2025-candidate + main. spec.diff,
  spec.history, and edition-pinned clause.get now work across the
  402 annual line.
• ECMA-402 proposals are indexed. proposal.list / proposal.get
  now cover the ecma402/ proposal set (~32 proposals that were
  entirely missing), and proposal.list gains a spec filter
  ({ spec: "402" }). Every proposal row carries a spec tag.

Removed

• Dropped the transient es2025-candidate 402 pin. It predated the
  final es2025 branch, which now supersedes it. clause.get,
  spec.diff, and the other edition-aware tools no longer accept
  edition: "es2025-candidate" — use es2025.

Changed

• Snapshots are sourced through a cache → hosted Worker → bundled
  fallback chain (loadSnapshot). The stdio server caches each
  snapshot under ~/.cache/tc39-mcp/ on first network fetch and serves
  it from disk thereafter, revalidating against the Worker only after a
  ~4-hour freshness window (conditional If-None-Match).
• latest on ECMA-402 now resolves to es2025 (the newest annual
  edition), matching 262's "latest = newest stable" semantics. Use
  main / draft / next for the working draft.
• The npm tarball shrinks ~70%. It now bundles only the offline
  fallback — spec-262-es2025, spec-262-main, spec-402-es2025,
  spec-402-main, and the proposals + test262 indexes — instead of
  every parsed edition. All other editions are fetched from the Worker
  on demand.
• cheerio and @tc39/ecma262-biblio move to devDependencies.
  They're only used by the parser at build time (npm run parse); the
  running server reads pre-parsed JSON and never imports them. A
  consumer install no longer pulls cheerio's ~20-package HTML-parsing
  tree (~2.4 MB). The published tarball also drops dist/docs and
  dist/refresh (build/CI-only code that runs from src/ via tsx,
  never at runtime).
• Refresh decouples from the npm version. refresh.yml updates R2
  every ~4 hours (live freshness for networked users) and re-bakes the
  npm bundle at most monthly, ending the ~2000-PATCH-bumps-per-year
  flood. A new annual edition still publishes immediately via the
  normal code-release path.
• @tc39/ecma262-biblio is pinned to an exact version (per its
  README's "pin a precise version" guidance), so the parse metadata
  layer is reproducible alongside the SHA-pinned spec HTML.
• The 262 parser gained an HTML-discovery fallback: since the
  pinned biblio is a snapshot of main, it can lag the HTML being
  parsed (a newer main, or an older edition carrying clauses since
  dropped). Any <emu-clause> the biblio doesn't list is now still
  captured, with metadata synthesized from the element — so a stale or
  mismatched biblio can no longer silently drop a clause.

Fixed

• Conditional revalidation returns 304 from the hosted Worker. The
  stdio loader sends If-None-Match when re-checking a live snapshot
  past its ~4-hour freshness window; the /r2/ proxy previously ignored
  it and re-sent the full body. It now returns a bodyless 304 on an
  etag match, so a revalidation costs a header round-trip instead of a
  tens-of-MB re-download.

Notes

• stdio now makes HTTPS requests to source non-cached snapshots
  (by default the hosted Worker; override with TC39_MCP_BASE_URL to
  self-host or air-gap). Tool arguments and clause ids never leave the
  process — see docs/privacy.md.
• v0.1.x installs are unaffected; this is a backward-compatible minor
  at the tool/protocol level.