Arbor v2.4.0 — The Agent-Native Leap
Release date: July 2026
First code-graph MCP server built for MCP 2026-07-28.
What's New
MCP 2026-07-28 Protocol
- Protocol version
2026-07-28with dual-version fallback for2025-03-26clients - Stateless
server/discoverendpoint - Response caching metadata (
ttlMs,cacheScope) on list/read operations - Extensions capability map for Tasks and MCP Apps
Tasks Extension
tasks/get,tasks/update,tasks/cancelfor long-running operations- Background indexing returns task handles instead of errors during cold start
- Agents poll task progress while the graph builds
MCP Apps (SEP-1865)
- Interactive blast-radius graph UI (
ui://arbor/blast-radius) - Interactive architecture map UI (
ui://arbor/architecture-map) analyze_impactandget_architecture_overviewdeclare_meta.uiresource URIs
Streamable HTTP Transport
arbor bridge --http --port 3333- Stateless HTTP alongside stdio
Mcp-Method/Mcp-Nameheader routing- Deploy behind load balancers for remote/enterprise use
Tool Quality
get_blast_radius— real git-diff blast radius analysis (replaces stub)- Pagination on
search_symbolsandget_map(offset,limit,hasMore) - Async tokio stdio (replaces blocking stdin loop)
Benchmarks
- Criterion suite:
cargo bench -p arbor-graph - CI regression gate via
.github/workflows/benchmarks.yml - Updated
docs/BENCHMARKS.mdwith token-savings methodology
Upgrade Guide
MCP clients (Claude / Cursor) — no changes required for stdio:
claude mcp add --transport stdio --scope project arbor -- arbor bridgeHTTP clients (new):
arbor bridge --http --port 3333
# POST http://127.0.0.1:3333/mcp
# Headers: Mcp-Method: tools/call, Mcp-Name: analyze_impactClients on 2025-03-26 keep working. Clients on 2026-07-28 get Tasks, Apps, caching, and pagination.
Full Changelog
Added
- MCP 2026-07-28 protocol:
server/discover,_metaparsing, response caching, dual-version fallback - Tasks extension:
tasks/get,tasks/update,tasks/cancel - MCP Apps: blast-radius graph + architecture map HTML templates
- Streamable HTTP:
arbor bridge --http [--port 3333] - Real
get_blast_radiusvia sharedarbor-graph::compute_blast_radius - Pagination on
search_symbolsandget_map - Criterion benchmarks + CI workflow
Changed
- Async MCP stdio (tokio replaces blocking stdin loop)
- MCP tool annotations:
_meta.uionanalyze_impactandget_architecture_overview - Workspace version bumped to 2.4.0
Deferred to v2.5.0
- Parallel (rayon) indexing
- Incremental PageRank
- Unified parser pipelines
- Process-level graph daemon
Install: cargo install arbor-graph-cli · GitHub · MCP Registry
🤖 Arbor MCP Quick Install
Use Arbor as a local MCP server in your AI client:
Claude Code (project-scoped)
claude mcp add --transport stdio --scope project arbor -- arbor bridge
claude mcp listIn Claude Code, run:
/mcp
Cursor MCP config (.cursor/mcp.json)
{
"mcpServers": {
"arbor": {
"command": "arbor",
"args": ["bridge"]
}
}
}VS Code MCP config (.vscode/mcp.json)
{
"servers": {
"arbor": {
"type": "stdio",
"command": "arbor",
"args": ["bridge"]
}
},
"inputs": []
}MCP Directories
- Glama: https://glama.ai/mcp/servers/@Anandb71/arbor
- Skills Playground: https://skillsplayground.com/mcps/nandb71-arbor/
Arbor v2.4.0
Installation
# Cargo
cargo install arbor-graph-cli --version 2.4.0
# Homebrew (macOS/Linux)
brew install Anandb71/tap/arbor
# Scoop (Windows)
scoop bucket add arbor https://github.com/Anandb71/arbor
scoop install arbor
# npm
npx @anandb71/arbor-cli
# Docker
docker pull ghcr.io/anandb71/arbor:v2.4.0Assets
| Platform | Architecture | File |
|---|---|---|
| Linux | x86_64 | arbor-linux-x86_64.tar.gz |
| Linux | aarch64 | arbor-linux-aarch64.tar.gz |
| macOS | x86_64 | arbor-macos-x86_64.tar.gz |
| macOS | Apple Silicon | arbor-macos-aarch64.tar.gz |
| Windows | x86_64 | arbor-windows-x86_64.zip |
What's Changed
Full Changelog: v2.3.0...v2.4.0