nikolai-vysotskyi/trace-mcp

v1.37.0 Breaking

This release includes breaking changes for platform teams planning a safe upgrade.

Published 19d MCP Developer Tools

View tool

✓ No known CVEs patched

Read the diff → Tool health → What is this tool? →

✓ No known CVEs patched in this version

Topics

ai-agents claude claude-ai claude-code claude-code-plugin claude-code-skill

+14 more

claude-desktop claude-skill claude-skills code-intelligence codex codex-skill developer-tools knowledge-graph mcp mcp-server llm token token-savings tokens

ReleasePort's take

Light signal

editorial:auto 9d

Release v1.37.0 introduces the toon output format for MCP responses, cutting token usage by 30‑60%.

Why it matters: The new toon format reduces response size (30‑60% fewer tokens), directly lowering bandwidth costs and improving latency for clients consuming MCP data.

Summary

AI summary

Added toon output format for MCP responses reducing token usage.

Changes in this release

Type	Severity	Summary	CVE
Feature	Medium	tools: toon output format for tabular MCP responses (30-60% fewer tokens) tools: toon output format for tabular MCP responses (30-60% fewer tokens) Source: llm_adapter@2026-05-21 Confidence: high	—
Dependency	Medium	Upgrade trace-mcp to latest version via npm install and init command Upgrade trace-mcp to latest version via npm install and init command Source: llm_adapter@2026-05-21 Confidence: low	—
Bugfix
Bugfix	Medium	daemon: cap remaining unbounded long-lived caches and stores daemon: cap remaining unbounded long-lived caches and stores Source: llm_adapter@2026-05-21 Confidence: high	—
Bugfix	Medium	daemon: plug FileWatcher re-entry leak + pass_cache TTL + progress-throttle pruning daemon: plug FileWatcher re-entry leak + pass_cache TTL + progress-throttle pruning Source: llm_adapter@2026-05-21 Confidence: high	—
Bugfix	Medium	daemon: restore /mcp routing for stdio clients + scope teardown to removed projects daemon: restore /mcp routing for stdio clients + scope teardown to removed projects Source: llm_adapter@2026-05-21 Confidence: high	—
Bugfix	Medium	pipeline: bound per-project resource accumulation across project lifecycle pipeline: bound per-project resource accumulation across project lifecycle Source: llm_adapter@2026-05-21 Confidence: high	—
Bugfix	Medium	telemetry: bound OTLP/Langfuse sink memory under unreachable export endpoint telemetry: bound OTLP/Langfuse sink memory under unreachable export endpoint Source: llm_adapter@2026-05-21 Confidence: high	—
Bugfix	Medium	analysis: restore findDuplicateSymbols helper that check_duplication relied on analysis: restore findDuplicateSymbols helper that check_duplication relied on Source: llm_adapter@2026-05-21 Confidence: low	—
Bugfix	Medium	benchmark: honest synthetic-estimator labelling + tokenizer calibration + multi-sample variance benchmark: honest synthetic-estimator labelling + tokenizer calibration + multi-sample variance Source: llm_adapter@2026-05-21 Confidence: low	—
Bugfix	Medium	db: add v11 domain tables to fresh-DB DDL db: add v11 domain tables to fresh-DB DDL Source: llm_adapter@2026-05-21 Confidence: low	—
Bugfix	Medium	db: mirror remaining v11+ migration tables in fresh-DB DDL + parity guard db: mirror remaining v11+ migration tables in fresh-DB DDL + parity guard Source: llm_adapter@2026-05-21 Confidence: low	—
Bugfix	Medium	intent: add missing domains JOINs in getCrossDomainDependencies intent: add missing domains JOINs in getCrossDomainDependencies Source: llm_adapter@2026-05-21 Confidence: low	—
Bugfix	Medium	scoring: drop stale-prone in-process cache from ranking_pins lookups scoring: drop stale-prone in-process cache from ranking_pins lookups Source: llm_adapter@2026-05-21 Confidence: low	—
Bugfix	Medium	test: bump bundles searchBundles limit-case timeout to 30s test: bump bundles searchBundles limit-case timeout to 30s Source: llm_adapter@2026-05-21 Confidence: low	—
Bugfix	Medium	tools: invert source/target check in get_cross_workspace_impact tools: invert source/target check in get_cross_workspace_impact Source: llm_adapter@2026-05-21 Confidence: low	—
Bugfix	Medium	topology: apply contractType filter to endpoints in get_api_contract topology: apply contractType filter to endpoints in get_api_contract Source: llm_adapter@2026-05-21 Confidence: low	—
Bugfix	Medium	topology: replace stale fr.* alias with sp.* in client-call queries topology: replace stale fr.* alias with sp.* in client-call queries Source: llm_adapter@2026-05-21 Confidence: low	—

Full changelog

1.37.0 (2026-05-15)

Features

tools: toon output format for tabular MCP responses (30-60% fewer tokens) (2e359ba)

Bug Fixes

analysis: restore findDuplicateSymbols helper that check_duplication relied on (8d49139)
benchmark: honest synthetic-estimator labelling + tokenizer calibration + multi-sample variance (279f619)
daemon: cap remaining unbounded long-lived caches and stores (f7b7044)
daemon: plug FileWatcher re-entry leak + pass_cache TTL + progress-throttle pruning (cc4f155)
daemon: restore /mcp routing for stdio clients + scope teardown to removed projects (4a708a2)
db: add v11 domain tables to fresh-DB DDL (12e4e05)
db: mirror remaining v11+ migration tables in fresh-DB DDL + parity guard (c0c052d)
intent: add missing domains JOINs in getCrossDomainDependencies (b39b5b3)
pipeline: bound per-project resource accumulation across project lifecycle (61fcd70)
scoring: drop stale-prone in-process cache from ranking_pins lookups (d5de560)
telemetry: bound OTLP/Langfuse sink memory under unreachable export endpoint (20397a4)
test: bump bundles searchBundles limit-case timeout to 30s (06e0df5)
tools: invert source/target check in get_cross_workspace_impact (5597781)
topology: apply contractType filter to endpoints in get_api_contract (db91b91)
topology: replace stale fr.* alias with sp.* in client-call queries (f304139)

Tests

tools: behavioural coverage for api-contract and service tools (f8c8872)
tools: behavioural coverage for compare and suggest tools (61b07cf)
tools: behavioural coverage for context and complexity tools (8334270)
tools: behavioural coverage for corpus and packaging tools (c23eb2d)
tools: behavioural coverage for cross-domain and edge tools (7557689)
tools: behavioural coverage for decision lifecycle tools (fd102ae)
tools: behavioural coverage for graph-traversal and analysis tools (1ca1b1a)
tools: behavioural coverage for graph/relationship tools (ac22e25)
tools: behavioural coverage for health and risk tools (acfb835)
tools: behavioural coverage for history and coupling tools (30b46fa)
tools: behavioural coverage for index, embedding, and perf tools (330a20c)
tools: behavioural coverage for ownership and tech-debt tools (0ce20be)
tools: behavioural coverage for quality and security tools (9987aed)
tools: behavioural coverage for quality/coverage tools (d6d6d26)
tools: behavioural coverage for refactoring tools (347d107)
tools: behavioural coverage for security and refactor tools (68d6171)
tools: behavioural coverage for session and analytics tools (fdc0288)
tools: behavioural coverage for session and analytics tools (a1d87ba)
tools: behavioural coverage for snapshot and maintenance tools (eb949d7)
tools: behavioural coverage for subproject and api-surface tools (83c3300)
tools: behavioural coverage for trend and visualization tools (2fcae49)
tools: behavioural coverage for workspace and architecture tools (8496a18)
tools: behavioural test coverage for search/get_outline/find_usages/get_change_impact/pin/query_decisions (2379ed1)
tools: skip discover_claude_sessions fixture on Windows (6ea757e)

Upgrade

npm install -g trace-mcp@latest
trace-mcp init

View diff on GitHub

Weekly OSS security release digest.

The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.

No spam, unsubscribe anytime.

Share this release

Share on X Share on Bluesky

Track nikolai-vysotskyi/trace-mcp

Get notified when new releases ship.

About nikolai-vysotskyi/trace-mcp

Framework-aware code intelligence that indexes source code into a cross-language dependency graph. Understands framework semantics — routes, ORM relations, component rendering, DI trees — for navigation, impact analysis, call graphs, refactoring, security scanning, and cross-session memory.