Repository navigation
Conversation
The official MCP Registry accepts a server only when the npm package it names carries mcpName. pack-trail.mjs now puts mcpName io.github.trailhq/trail-code on @trailhq/trail-code alone, and both packages get MCP keywords for npm search. server.json is the registry entry (from mcp-publisher init, then fixed: description under the 100-character limit, the trailhq/Trail-Code URL, the `mcp` argument so npx starts the server rather than the CLI, and no placeholder API key). .mcp.json at the root is how Cursor Directory and others find the server. The repo's package.json keeps the @nanonets/graft name: the registry reads npm, not the repo, and tests assume this checkout runs as graft.
… keeps the graft name
…d:org token, not the device login
Anirudh's #723, #724, #730, #734, #736, #737 and #740 (share-session still behind TRAIL_SHARE_SESSION=1), #707's WASM depth tier, and the 0.28.1 MCP listing from #742, which never reached main. trail check flags CHANGELOG.md against two learnings on sign-in testing and single claude -p runs. Those are word matches on the changelog's wording; the release only describes code already merged.
# Conflicts: # CHANGELOG.md
Everything merged since 0.29.0: Anirudh's #725-#733, #747-#750 and #828, and the community batch (new languages, Pi host, resolver and Windows fixes). Also carries release/0.29.0 (#761), which never reached main, so main gets the 0.29.0 changelog and the MCP listing too. share-session stays behind TRAIL_SHARE_SESSION=1. trail check flags CHANGELOG.md against two learnings on sandbox PATHs and single claude -p runs. Those are word matches on the changelog's wording; the release only describes code already merged.
🌱 graft blast radiusNothing outside this diff depends on it. 1 area changed; no indexed dependents at depth 2. Who knows this code — 1 person across 1 area
Ownership is git history over each area's own files, weighted towards recent work (120-day half-life). Merge commits and bots are dropped, and you are dropped from your own PR. A name with no Test signal per changed area — 1 –Reached = a node under a test path has a resolved edge into the changed symbol. It undercounts anything called indirectly — through a CLI, a spawned process or a dynamic import — so read a low ratio as “look here”, never as a coverage gate.
Open the interactive graph → — click an area to see its dependent symbols at file:line. |
Bumps both packages to 0.30.0 with everything merged since 0.29.0 (71 commits). It also carries #761 (the 0.29.0 release branch), which never merged into main, so main finally gets the 0.29.0 changelog and the MCP listing. Merging this one makes #761 and #742 redundant.
What's in it
trail init(Let one developer use trail just for themselves #750), Claude Desktop and Codex (Make trail work in Claude Desktop and Codex #733), stale learnings flagged (Flag learnings that name code that's gone #731), Claude memory import (Import a Claude memory export in trail init #732), one learning file per session (One learning file per session, or one per PR #748), the terminal title keeps the repo name (Stop trail hiding which repo a terminal is in #749), share sync re-sends only grown sessions (Share sync re-sends only sessions that grew #828), TS barrel files (Connect TS barrel files and generic supertypes in the graph #726), the tokens-saved counter (Keep the tokens-saved counter moving in long sessions #747), and telemetry send time (Stamp telemetry events with the time they're sent #728). Also Commit trail's shared setup in Trail-Code #725 and Keep the sources-diff summary to one line #729.aliascalls, the Pi host, more callers resolved in TS/Go/Kotlin/Python/PHP, Windows console and blast-radius fixes,--strictbuilds, and--include-dirfor dot-folders.share-sessionis still behindTRAIL_SHARE_SESSION=1. Nobody removed the check, so it's left as it was.Checked on the packed tarballs
mcpNameis on trail-code only.npm i -gof both installs 56 packages with no.nodefiles, so nothing compiles.🤖 Generated with Claude Code