Agent Lsp logo

Agent Lsp

by blackwell-systems

Orchestrates real language servers for agent-driven code navigation, impact analysis, diagnostics, and speculative edits.

Install Agent Lsp

Pick your client, then add it in one click or paste the config.

STDIOPython
~/Library/Application Support/Claude/claude_desktop_config.json
claude_desktop_config.json
{
  "mcpServers": {
    "blackwell-systems-agent-lsp": {
      "command": "uvx",
      "args": [
        "agent-lsp"
      ]
    }
  }
}

Settings → Developer → Edit Config, merge this into mcpServers, then fully quit and reopen Claude Desktop.

We couldn’t confirm it starts correctly

We ran the install command but it didn't respond within our test window — this can mean a slow first-time install rather than a real problem.

uvx agent-lsp

No response to initialize.

This is an experimental automated check and can have false negatives — missing environment variables, a slow cold install, etc. It doesn’t necessarily mean something’s wrong. Last checked .

Overview

blackwell-systems/agent-lsp MCP server orchestrates existing language servers such as gopls, rust-analyzer, and jdtls, exposing their code intelligence through MCP. It maintains warm workspace sessions, supports cross-repository analysis, and groups multi-step operations such as blast-radius checks and caller discovery into agent-oriented workflows. Speculative sessions let agents evaluate diagnostics for edits before committing changes to disk. Reach for it when an agent needs symbol-aware refactoring, language-server diagnostics, or code navigation across substantial or multi-language repositories.

Use cases

  • Analyze refactor impact across callers and repositories
  • Preview code edits before writing them to disk
  • Navigate unfamiliar code with symbol-aware queries
  • Check diagnostics after simulated changes
  • Automate multi-language code workflows

Key features

  • Orchestrates existing LSP servers through MCP
  • Persistent warm workspace sessions
  • Blast-radius and caller analysis
  • Speculative edit sessions with diagnostic comparison
  • Cross-repository reference discovery
  • GCF-encoded tool responses with JSON fallback

Tools65

What Agent Lsp lets your agent do. Expand a tool for its parameters.

Verified live~12.9k tokens of context

Security & health

What it can reach and how reliably it runs — from our own checks, not the maintainer’s claims.

ActiveLast checked

Recent health check succeeded.

30/30 checks healthy over the last 97d

Supply-chain signal

No high-severity advisories surfaced by our automated scan.

Critical 0High 0Medium 0Low 0

Scanned 10d ago via OSV.dev · agent-lsp (PyPI)

Adoption & maintenance

Factual signals from GitHub, npm, and our automated checks — not a rating.

GitHub stars180Stargazers on the source repository.
Last commit1d agoMost recent push to the default branch.
Install checkInconclusiveDidn't respond in our test window — often a slow first install.
Tools exposed65Callable tools this server registers over MCP.
Directory activity18 viewsConfig copies, upvotes, and views on AllMCPs.

How Agent Lsp works

What blackwell-systems/agent-lsp does

blackwell-systems/agent-lsp acts as an orchestration layer between AI agents and existing Language Server Protocol implementations. It is not itself a language server: it starts and coordinates servers such as gopls, rust-analyzer, and other language-specific implementations, then exposes their capabilities through MCP tools.

The server is intended for code intelligence rather than plain text search. Agents can locate symbols, inspect callers and references, retrieve completions and diagnostics, inspect semantic tokens and inlay hints, and request formatting edits. Workspace-level operations include exported-symbol impact analysis, test versus non-test caller partitioning, cross-repository references, and change-risk classification.

How it works

A single Go binary routes files to language servers based on their languages and keeps runtime state across files and projects. Persistent sessions allow indexes to remain warm instead of rebuilding for every request. start_lsp initializes the language-server runtime, while add_workspace_folder adds repositories to a multi-root workspace when the underlying server supports that model.

For safer modifications, an agent can create a simulation session, make in-memory changes, compare diagnostics with the baseline, and either commit or discard the session. commit_session can write changes to disk or return a unified diff, while destroy_session releases the session resources. Direct edits can be applied with a workspace edit or a text-match replacement; the documented workflow calls for previewing an edit before applying it.

The server also supports skill-based phase enforcement. An agent can activate a skill workflow so tool calls are checked against permitted phases, then deactivate it when the workflow ends.

Setup and configuration

Install the binary and initialize it with:

Terminal
curl -fsSL https://raw.githubusercontent.com/blackwell-systems/agent-lsp/main/install.sh | sh && agent-lsp init

The language servers used for a project need to be available to the agent-lsp runtime. detect_lsp_servers scans a workspace for source languages, checks PATH for matching server binaries, and returns a suggested configuration. Use start_lsp before workspace-folder operations such as add_workspace_folder or cache export.

Responses use GCF by default. Set AGENT_LSP_OUTPUT_FORMAT=json to switch output back to JSON.

Tools and capabilities

  • Find symbols, references, callers, document highlights, completions, inlay hints, semantic tokens, and diagnostics through LSP operations.
  • Inspect a symbol with combined type, source, caller, reference, and test-caller information.
  • Calculate blast radius, including transitive callers and test/non-test partitions.
  • Analyze cross-repository references by adding consumer repositories to the workspace.
  • Preview, simulate, evaluate, apply, commit, or discard code changes.
  • Format complete documents or selected ranges, and execute commands returned by language-server code actions.
  • Detect workspace languages and installed LSP binaries, refresh watched-file state, and export a compressed reference cache for team sharing.

The project reports CI verification across 32 languages and describes 24 agent workflows, including refactoring, safe editing, and verification pipelines.

Limitations and notes

Results depend on the capabilities and availability of the language servers configured for the workspace. Some operations are language-server-specific; for example, multi-root indexing requires support from the selected server. Open documents and simulation sessions consume resources, so close_document and session-destruction operations are relevant in long-running or large-codebase sessions. GCF can be disabled when a client or processing pipeline requires JSON output.

Reviews

No reviews yet — be the first to share how it worked for you.

Frequently Asked Questions about Agent Lsp

blackwell-systems/agent-lsp is an MCP server that connects AI agents to existing language servers such as gopls, rust-analyzer, and jdtls. Its main tools provide symbol search, references, callers, diagnostics, formatting, blast-radius analysis, cross-repository navigation, and speculative edit sessions before changes are committed.

Install it with the project’s shell command, then run `agent-lsp init`. The command downloads the single Go binary; `agent-lsp init` initializes the setup. Use `detect_lsp_servers` afterward to inspect workspace languages, find language-server binaries on `PATH`, and generate suggested configuration.

Create a simulation session, make the proposed edits in memory, and evaluate the resulting diagnostics before committing. Use the preview workflow before `apply_edit`; then use `commit_session` to write changes or return a unified diff, or discard and destroy the session if the result is unsafe.

Yes, it can analyze multiple repositories in one workspace when the selected language server supports multi-root workspaces. Call `start_lsp`, add repositories with `add_workspace_folder`, and use `get_cross_repo_references` to separate references in the primary library from references in consumer repositories.

Set `AGENT_LSP_OUTPUT_FORMAT=json` before running the server. GCF is the default response format, and this environment variable switches tool output back to JSON for clients or processing pipelines that require it.

Related MCP servers

Add a badge to your READMEShow it’s listed — live SVG, Markdown or HTML.

AllMCPs Directory Badge

Full Badge Customizer

Showcase your server listing on GitHub or your project documentation. Embed this dynamic SVG badge to highlight official listing status and live engagement.

Badge Style:
Live Dynamic SVG PreviewAgent Lsp AllMCPs Directory Badge
Markdown (GitHub README)
[![AllMCPs](https://allmcps.com/api/badge/blackwell-systems-agent-lsp?style=directory)](https://allmcps.com/mcp/blackwell-systems-agent-lsp)
HTML Embed
<a href="https://allmcps.com/mcp/blackwell-systems-agent-lsp"><img src="https://allmcps.com/api/badge/blackwell-systems-agent-lsp?style=directory" alt="Agent Lsp on AllMCPs" /></a>