Releases: LonelyTraderBay/memory-for-ai
Release list
v0.11.0
Highlights
Symbol-level editing is here — the server grows from 18 to 22 MCP tools.
edit_symbol— edit a symbol's source using its graph location (replace_body/insert_before/insert_after). Dry-run by default: returns a diff plan and writes nothing untildry_run=false. Pre-write guards reject stale index ranges, nested-symbol destruction (withoutforce), and externally modified files.delete_symbol— delete a symbol with a graph-based usage check. Refused while the symbol has direct callers (the refusal lists them); the dry-run plan shows removed lines, callers, and orphan candidates.rename_symbol— rename with a mandatory coverage check so renamed symbols stay resolvable.undo_edit— restore any file from the newest edit-tool backup. The pre-undo content is itself backed up, so undo is undoable.
Every successful write keeps a backup under the cache dir's backups/ and re-indexes the project incrementally.
Also in this release
- Docs: full 22-tool sync across README, AGENT_GUIDE, INSTALL, llms.txt, product-metadata, landing page; the fixed tool-manifest cost is now ~9K tokens/session (measured).
- Docs: A/B self-measurement report on this repo (8 questions, graph 8/8 PASS at ~3.4K est. tokens vs ~765K file-by-file).
- Windows: bundled MinGW toolchain setup script for building from source without a system compiler.
- Tests: new
editsuite, 36 tests, 143/143 suites green (7,720 assertions).
Full Changelog: v0.10.9...v0.11.0
v0.10.9
Full Changelog: v0.10.8...v0.10.9
Security Verification
Every binary published below was scanned by VirusTotal before smoke and soak testing, and the verdict for the exact shipped bytes is linked per product (decisive engines: 59–68).
1 shipped binary/binaries carried only the documented single Microsoft machine-learning !ml result; no other decisive engine reported malicious or suspicious.
| Product | Shipped binary | VirusTotal verdict |
|---|---|---|
linux-amd64 |
24119df0c47ff147367340b624ac6a7c490e2348434a3f36d37ca02900322ec4 |
clean |
linux-arm64 |
eb36134a38ede21f729d8efd28424f82ee949cc5326c0875bb0cc05a85f6f412 |
clean |
linux-amd64-portable |
26920af824ef2ebb989d1822291e7772934089eaa3c96cac62012f554be4bc05 |
Microsoft !ml |
linux-arm64-portable |
7c595fa5d5100d01e6bb8c582abbd98f1463e013b698fc83fe2fe23d4b3d08d9 |
clean |
darwin-amd64 |
d56a24980443a187381f188762409e4e3d78e20c1b0b78ac5afbeaf12d577942 |
clean |
darwin-arm64 |
67c95e36586cb0e119b58974436625af17fe7a64a6cb3db650600ceffd196ab3 |
clean |
windows-amd64 |
0221ffef33da4bb88c4fa98c4abb7ed57044dbb112dea5685904eae4a76efb5a |
clean |
windows-arm64 |
99ecdcc4b1d78ffcbc806edb504d94b97792accd217f7a64911276bca9e46179 |
clean |
Selection is tuple-local and defaults to stripped. The selected executable SHA-256 was verified again after packaging; archive containers were not redundantly submitted to VirusTotal. Their hashes remain available in checksums.txt.
Durable evidence: candidate provenance, candidate VirusTotal results, selection decisions.
v0.10.8
First public release.
memory-for-ai is an MCP server that turns a codebase into a persistent knowledge graph — functions, classes, call chains, HTTP routes, cross-service links — so an AI coding agent answers structural questions with graph queries instead of reading file after file.
- One self-contained native executable; 162 languages via vendored tree-sitter grammars, refined by embedded Hybrid-LSP type resolution (full type-aware resolution for Python, TypeScript/JavaScript/JSX/TSX, PHP, C#, Go, C/C++, Java, Kotlin, Rust, Perl; syntactic fallback for every other grammar).
- 18 MCP tools, each also runnable as a local one-shot CLI command.
- Per-project install (
install --project): a fenced, repo-named server per repository, pinned with--scopeso one project's agent can never query another project's graph. - Auto-configures 45 supported agent surfaces, auto-indexes on first use, and keeps the graph fresh with a background watcher.
- 100% local: no telemetry, nothing phones home, no API key, no Docker, no language runtime required.
Install: one line for macOS/Linux, PowerShell for Windows, plus npm/PyPI/Homebrew/Scoop/Winget/Chocolatey/AUR/Go/Nix — see the README quick start.
Every release archive is SHA-256 verified against checksums.txt, SLSA Level 3 attested, and cosign-signed; VirusTotal-scanned before publication. Verification steps: SECURITY.md.