Release / / with 0.6.0, 0.6.1, 0.7.0, 0.8.0
codebase-memory-mcp v0.8.1
Find code by meaning as well as by names and literal text.
Before you upgrade to v0.8.1
1 to fixNim parsing is removed in v0.8.1. Use another analyzer for Nim; no replacement grammar is stated.
Read the evidence
Type-aware resolution expands to Python, PHP, TypeScript/JavaScript, C#, Java, Kotlin and Rust, alongside Go and C/C++.
Type-aware resolution runs in all indexing modes; `mode="advanced"` now falls back to `full`.
Read the evidence
The default search result limit is capped at 200 instead of allowing the earlier very large default.
Read the evidence
Cluster summaries now expose multi-level Leiden communities instead of single-level Louvain groups.
Read the evidence
Queries gain `EXISTS {}`, `COUNT(DISTINCT)`, label alternation and more scalar functions; unsupported syntax now raises errors instead of silently returning empty results.
Read the evidence
Invalid regular expressions now return errors; a wildcard-free `file_pattern` acts as a substring filter.
Read the evidence
`CBM_CACHE_DIR` selects the database directory; `CBM_WORKERS` overrides worker count with container-aware resource detection.
Helm charts gain template calls and chart dependency edges; HCL block labels are included in node names.
Read the evidence
Service detection expands to gRPC, GraphQL and tRPC, including protobuf routes and gRPC stub calls.
Read the evidence
Setup adds Kiro CLI and an `install --plan` preview, honors `CLAUDE_CONFIG_DIR`, and supports fish PATH setup.
The advertised tracing tool is `trace_path`; `trace_call_path` remains an accepted alias.
Read the evidence
Indexing writes a compressed graph artifact and adds a `.gitattributes` merge rule; committing the artifact remains optional.
Read the evidence
What changed in v0.8.1
The 0.6–0.8 release pages, published together on June 12, add semantic and BM25 search, graph artifacts, cross-repository relationships and expanded query syntax. v0.8.1 removes Nim parsing.
Also: Shared graph snapshots, cross-repository links, more language coverage and type-aware call resolution.
v0.8.1; older 0.6.0–0.8.0 release pages are archived without binaries. Native and portable Linux distributions are documented, alongside npm and PyPI installation.
New things you can do
Find code using semantic keywords or full-text ranking.
Read the evidence
Find structurally similar functions and semantic relationships in the graph.
Read the evidence
Bootstrap a teammate's local graph from a compressed snapshot stored with the repository.
Read the evidence
Explore dependency relationships across multiple indexed repositories.
Read the evidence
Inspect argument-to-parameter data flow and publish-subscribe channel relationships.
Read the evidence
Find functions with high complexity, deep loops or scan patterns through graph queries.
Read the evidence
Build and run local code exploration on BSD systems.
Read the evidence
Fixed and improved
Degree and entry-point filters and `include_connected` now apply instead of being silently ignored.
Read the evidence
The `since` parameter now participates in Git change detection.
Read the evidence
Architecture aspects filtering is applied and the computed overview is connected to the MCP handler.
Code search uses PowerShell and Windows paths containing non-ASCII characters are handled with wide-character APIs.
Nested ignore rules, Git submodule dirty-state detection and moderate-mode subtree handling are corrected; excluded directories are reported.
Complex C++ indexing no longer hits the documented template-resolution crashes; a corrupt cache no longer crashes every project session.
Read the evidence
Seeded by @iluxav with Codex, from the release's public sources. Each change links to its evidence.