Skip to content

blackwell-systems/agent-lsp

v0.10.0 Feature

This release adds 2 notable features for engineering teams evaluating rollout.

✓ No known CVEs patched
Read the diff → Tool health → What is this tool? →

✓ No known CVEs patched in this version

Topics

agentskills ai-agents ai-tooling claude claude-code code-intelligence
+13 more
code-quality coding-agent developer-tools go language-server-protocol lsp mcp mcp-server model-context-protocol python rust static-analysis typescript

Summary

AI summary

get_change_impact now includes methods, per-symbol test callers, and other agent‑evaluated quality improvements.

Full changelog

v0.10.0: Agent-Evaluated, Agent-Improved

61 tools. 30 languages. 23 skills. This release was shaped by running the same 10-task evaluation across four AI models (Claude Opus 4.6, Cursor, GPT-5.5, Gemini 2.5 Pro) and fixing every friction point they found. Four models, four independent evaluations, all recommend agent-lsp.

get_editing_context (tool #61)

One call returns everything an agent needs before editing a file: symbols with signatures, callers partitioned by test vs non-test, callees, and imports. Supports ETag for conditional responses and includes token savings metadata.

/lsp-onboard skill (23rd skill)

First-session project onboarding. Explores the codebase via LSP: detects languages and build system, identifies entry points, maps package structure, finds hotspots, checks pre-existing diagnostics. Produces a structured project profile for the session.

Agent-evaluated quality improvements

Four AI models independently evaluated agent-lsp and surfaced friction points. All fixed:

  • get_change_impact includes methods. Methods like (*Hub).SetSender were invisible because the receiver prefix ( failed the uppercase export check.
  • get_change_impact per-symbol test callers. New affected_symbols field nests test and non-test callers under each symbol.
  • position_pattern works without line/column. find_references and inspect_symbol now accept position_pattern alone.
  • preview_edit net_delta ignores hints. Severity 3/4 diagnostics no longer count.
  • destroy_session returns success for missing sessions. Returns "already_destroyed" instead of error.

Token savings metadata

list_symbols, get_symbol_source, and get_editing_context include _meta.token_savings showing tokens returned vs full file size.

ETag/conditional responses

File-scoped tools accept if_none_match. When the file hasn't changed, returns not_modified instead of recomputing.

Also

  • safe_delete_symbol column resolution fix (unexported symbols now work)
  • Flaky TestSubscribeHealth_Stop fixed
  • Token savings wired into handlers
  • 607 new test lines (coverage round 6)

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

Track blackwell-systems/agent-lsp

Get notified when new releases ship.

Sign up free

About blackwell-systems/agent-lsp

Stateful MCP server over real language servers. 50 tools, 30 CI-verified languages, 20 agent workflows. Persistent sessions keep the index warm across files and projects. Speculative execution simulates edits in memory before writing to disk.

All releases →

Related context

Earlier breaking changes

  • v0.11.0 `get_change_impact` renamed to `blast_radius`. Same handler, same parameters, new name.

Beta — feedback welcome: [email protected]