Blitz MCP server
Commit Nov 15, 2026Lets AI coding agents list collections, inspect request schemas, execute endpoints, and run API test suites via a built-in MCP server.
Also known as h3ma209/Blitz
- Stars
- 9
- Tools
- 37
- Downloads, 30 days
- —
- Latest version
- —
- Last commit
- Nov 15, 2026
- License
- —
What is Blitz MCP?
Blitz lets AI coding agents list collections, inspect request schemas, execute endpoints, and run API test suites via a built-in MCP server. It exposes 37 tools, such as blitz_list_collections, blitz_get_request, blitz_send_request.
Summary generated from the project's README
About Blitz MCP server
Developers using AI coding agents like Cursor, Claude, Antigravity, or Windsurf can use Blitz to interact with their APIs directly from a local Rust-based workbench. By providing a built-in MCP server with 37 tools, it allows these agents to list collections, inspect request schemas, execute endpoints, and run full API test suites via stdio. This setup works entirely locally, requiring no cloud accounts or logins, and keeps all collections and tokens on your machine.
Unlike heavy Electron-based alternatives, Blitz is a native desktop application that starts in under 150ms and uses approximately 40MB of RAM. It supports a wide range of protocols including REST, GraphQL, WebSocket, gRPC, MQTT, SSE, and Socket.IO. For security, it integrates with your OS keychain to protect secrets and uses Git-native synchronization so you can manage your API collections within your existing repositories using standard git commands.
What you can do with it
- 1Automated API testing with agentsAn AI agent can execute your existing API test suites to verify endpoint functionality.
- 2Schema inspection for AI contextYou can task an AI client with inspecting request schemas to understand how to interact with your services.
- 3Local protocol testingDevelopers can test various communication standards like WebSockets or gRPC through a single interface.
- 4Secure credential managementManage sensitive API tokens using your computer's native encrypted storage.
Written from the project's README. Every use case is checked against a quote from it.
Blitz MCP tools 37
Detected from the READMEblitz_list_collectionsblitz_get_requestblitz_send_requestblitz_execute_httpblitz_list_environmentsblitz_set_env_varblitz_set_active_environmentblitz_upsert_requestblitz_import_curlblitz_generate_codeblitz_list_historyblitz_run_collection
Show all 37 toolsShow fewer tools
blitz_suggest_assertionsblitz_perf_testblitz_sync_openapiblitz_get_collectionblitz_create_environmentblitz_delete_requestblitz_start_mockblitz_stop_mockblitz_export_collectionblitz_diff_responsesblitz_delete_collectionblitz_generate_docsblitz_create_folderblitz_delete_folderblitz_delete_environmentblitz_clear_historyblitz_import_harblitz_manage_cookiesblitz_websocket_sendblitz_grpc_unaryblitz_sse_listenblitz_mqtt_publishblitz_mqtt_subscribeblitz_graphql_ws_subscribeblitz_socketio_emit
Automated checks
Ran Nov 14, 2026Full report →Not verified yet. A server needs a score of 80, no serious finding, and enough data for us to check. How checks work · What to fix for more points
- Source repository linked and reachableh3ma209/blitz is public and reachablegithub.com/h3ma209/blitz ↗Pass
- Package names this repository as its sourceNot checked
- Build provenance (Sigstore attestation)Not checked
- Open source license (OSI)No licenseFinding
- Commit or release in the last yearLast commit or release 2026-11-10github.com/h3ma209/blitz ↗Pass
- OpenSSF ScorecardNot checked
- Package name not close to a popular packageNot checked
- No critical advisory in the package or its dependenciesNot checked
- No hidden instructions to the model37 tool descriptions and the README read; no override phrases, credential paths or hidden charactersPass
- Destructive tools are labelledNo tool named like delete, drop or execPass
- Remote endpoint answers the MCP handshakeNot checked
README
View on GitHub⚡ Blitz — AI-Native API Client with Built-in MCP Server
⚡ What is Blitz?
Blitz is a lightweight, open-source API development and testing client written in pure Rust (Dioxus). It replaces heavy Electron-based tools with a native desktop application that respects your hardware and privacy:
- 🤖 Built-in MCP server (37 tools): Cursor, Claude, Antigravity, and Windsurf can list collections, send requests, sync OpenAPI, run perf tests, and diff responses — locally, over stdio.
- 🔒 100% Offline & Local-First: No accounts, no logins, no cloud tracking. Your collections, tokens, and environments stay on your machine.
- 🐙 Git-Native Collaboration: No $19/user/month cloud workspace paywalls. Save collections right inside your Git repo and sync via
git pull/git push. - 🚀 Native speed: Starts in <150ms and sits at ~40MB RAM (CLI runner ~10MB). No bundled Chromium or multi-process Electron overhead.
- 🌐 All Major Protocols: REST, GraphQL, WebSocket, gRPC (Protobuf), SSE, Socket.IO, and MQTT in a unified, modern interface.
📸 Interface Preview
⚖️ Blitz vs Alternatives
| Feature | Postman | Bruno | Yaak | ⚡ Blitz |
|---|---|---|---|---|
| Language & Engine | Electron + Node.js | Electron + Node.js | Tauri + React | Pure Rust + Dioxus |
| Startup Time | ~3,500 ms | ~1,200 ms | ~450 ms | < 150 ms |
| Memory Footprint | 800 MB – 1.5 GB | 150 MB – 300 MB | ~120 MB | ~40 MB (GUI) / ~10 MB (CLI) |
| Account Required | Mandatory Cloud | No | No | Never (100% Offline) |
| Team Sync | Paid Cloud ($19+/mo) | Git (.bru files) | Git / Cloud | Git-Native (standard JSON) |
| AI Agent Bridge | Proprietary AI add-on | None | None | Built-in MCP Server (37 tools) |
| Protocols | REST / WS / gRPC | REST / GraphQL | REST / WS / gRPC | REST / WS / gRPC / MQTT / SSE |
| Secret Storage | Cloud or Plaintext | Plaintext .env |
SQLite | OS Encrypted Keychain |
| Mock Server | Cloud-only | No | No | Localhost 127.0.0.1 |
| License | Proprietary | MIT | Commercial / Fair-Core | MIT / Apache-2.0 |
📦 Downloads & Installation
⚡ 1-Line Quick Install
macOS
curl -fsSL https://raw.githubusercontent.com/h3ma209/Blitz/master/install.sh | bash
Downloads latest DMG, installs to /Applications/Blitz.app, and automatically clears Gatekeeper quarantine.
Linux
curl -fsSL https://raw.githubusercontent.com/h3ma209/Blitz/master/install.sh | bash
Downloads latest binary and installs executable to ~/.local/bin/blitz.
Windows
PowerShell:
irm https://raw.githubusercontent.com/h3ma209/Blitz/master/install.ps1 | iex
Command Prompt (curl):
curl.exe -fsSL https://raw.githubusercontent.com/h3ma209/Blitz/master/install.ps1 -o install.ps1 && powershell -ExecutionPolicy Bypass -File install.ps1
Git Bash / WSL:
curl -fsSL https://raw.githubusercontent.com/h3ma209/Blitz/master/install.sh | bash
Installs to %LOCALAPPDATA%\Blitz\blitz.exe, adds to User PATH, and creates Start Menu shortcut.
Homebrew Tap (recommended on macOS)
brew tap h3ma209/blitz https://github.com/h3ma209/homebrew-blitz
brew install --cask blitz
Windows package managers
After manifests are accepted upstream (templates in packaging/winget/ and packaging/scoop/):
winget install --id h3ma209.Blitz -e
scoop install blitz
Until then, use the 1-line PowerShell installer or Releases.
Direct Downloads
Pre-built binaries for macOS, Linux, and Windows are available directly from the Releases Page.
| Platform | Download | Package / Command |
|---|---|---|
| macOS (Apple Silicon) | Blitz-macos-arm64.dmg | Drag to Applications |
| macOS (Tarball) | blitz-macos-arm64.tar.gz | tar -xzf blitz-macos-arm64.tar.gz |
| Linux (x86_64) | blitz-linux-x86_64.tar.gz | tar -xzf blitz-linux-x86_64.tar.gz |
| Windows (x86_64) | blitz-windows-x86_64.zip | Unzip & run blitz.exe |
| Cargo (Any OS) | Source crate | cargo install --git https://github.com/h3ma209/Blitz.git |
🛡️ Install trust & OS prompts
Blitz is unsigned (no Apple notarization / Windows EV cert yet). Prefer Homebrew, winget, or Scoop when available. For how we handle secrets, MCP, and disclosures, see SECURITY.md.
Browser downloads get a quarantine flag. The install script clears it automatically; otherwise:
xattr -cr /Applications/Blitz.app
Or right-click Blitz in Applications → Open once.
Use More info → Run anyway, or the PowerShell installer (unblocks automatically), or:
Unblock-File "$env:LOCALAPPDATA\Blitz\blitz.exe"
chmod +x ~/.local/bin/blitz
🎯 Core Features & Capabilities
🔌 Multi-Protocol Support
Interact with any modern API standard in one place:
- REST / HTTP: Query parameters, headers, GraphQL, form-data, multipart uploads.
- WebSockets: Interactive bidirectional streaming, frame inspection, Socket Deck, and message presets.
- gRPC: Unary calls and streaming with Protobuf schema reflection and decoding.
- MQTT: Real-time publish and subscribe to IoT brokers.
- SSE (Server-Sent Events): Live streaming event reader.
- Socket.IO: Native engine handshake and event listener.
- MCP (Model Context Protocol): Direct connection to AI agent tooling and context servers.
📑 Comprehensive Collection Workspace
Manage complex API suites with dedicated tabs:
- Instant Search & Type Filtering: Real-time debounced sidebar search with zero UI lag. Type inline prefixes directly in the search bar to filter by node type:
folder:<query>orf:<query>(e.g.folder:authorfolder:alone to browse all folders)req:<query>orr:<query>(e.g.req:loginorreq:alone to browse all requests)col:<query>orc:<query>(e.g.col:store)- Alternative syntax:
is:folder,is:req,type:folder,type:req
- Overview: Collection documentation, structure preview, and real-time statistics.
- Authorization: Inherited authentication (Bearer, Basic, API Key, OAuth2) applied across child requests.
- Pre-request & Test Scripts: Collection-wide scripting for setup, teardown, and validation.
- Variables: Environment and collection variable scope resolution.
- Runner: Batch test execution with iteration controls, delays, and pass/fail diagnostics.
🧪 Postman-Compatible JavaScript Testing
Write familiar assertions and workflows locally without remote runtime limits:
blitz.test()andblitz.expect()assertion suites.blitz.sendRequest()for chained requests and token pre-fetching.- Built-in
CryptoJSlibrary (SHA256,HmacSHA256,MD5, Base64, Hex). - Dynamic variables (
{{$guid}},{{$timestamp}},{{var}}).
🔐 Secure Credential Storage
Protect sensitive environment secrets and tokens:
- Native OS Keychain integration (macOS Keychain, Linux Secret Service, Windows Credential Manager).
- Secrets stay encrypted locally and are automatically stripped before committing to Git remotes.
🔄 Seamless Migration & Local Mocks
- Postman: Import and export Postman v2.1 collections with full script fidelity.
- OpenAPI: Import and export OpenAPI 3.0 specifications.
- cURL & HAR: Paste raw terminal cURL commands or drop browser HAR files directly into the workbench.
- Local Mock Server: Instant zero-config HTTP mocks on
127.0.0.1served from your saved examples.
Quick Start
Prerequisites
- Rust Toolchain (1.83+)
- macOS, Linux, or Windows
Build and Run from Source
# Clone repository
git clone https://github.com/h3ma209/Blitz.git
cd Blitz
# Run debug build
cargo run
# Or run optimized release build
cargo run --release
Install as Native Desktop App
make install
| Operating System | Default Install Destination | Uninstall |
|---|---|---|
| macOS | ~/Applications/Blitz.app |
make uninstall |
| Linux | ~/.local/share/blitz/ + ~/.local/bin/blitz |
make uninstall |
| Windows | %LOCALAPPDATA%\Blitz\ |
powershell -File scripts/install.ps1 -Uninstall |
CI/CD Automation & AI Agent Integration
Headless CLI Runner
Run test collections headless in your CI/CD pipelines without cloud subscriptions or proprietary API keys:
# Execute collection headless (returns exit code 0 on pass, 1 on failure)
blitz run collection.json --env staging.json
🤖 AI Agent Integration via MCP (Cursor, Claude, Antigravity, Windsurf, VS Code, Zed)
Blitz includes a built-in Model Context Protocol (MCP) server (blitz mcp), allowing AI coding assistants and autonomous agents to list collections, inspect request schemas, execute endpoints, and run API test suites directly through Blitz.
Automatic IDE & Agent Setup:
Blitz can automatically scan your machine for installed IDEs and AI coding agents, and configure the MCP server for you:
# Scan and interactively select which IDEs/agents to configure:
blitz mcp setup
# Or configure all detected IDEs automatically without prompting:
blitz mcp setup --all
# Or target a specific IDE:
blitz mcp setup --target cursor
blitz mcp setup --target claude_desktop
blitz mcp setup --target vscode
blitz mcp setup --target zed
# Scan only (preview detected tools without modifying configs):
blitz mcp setup --scan
Supported IDEs and coding agents detected out of the box:
- Cursor (
~/.cursor/mcp.json) - Claude Desktop (
claude_desktop_config.json) - Antigravity / Gemini CLI (
~/.gemini/antigravity/mcp_config.json) - Windsurf (
~/.codeium/windsurf/mcp_config.json) - VS Code (
~/Library/Application Support/Code/User/mcp.json) - Cline extension for VS Code (
cline_mcp_settings.json) - Roo Code extension for VS Code (
cline_mcp_settings.json) - Zed (
settings.jsoncontext_servers) - Claude Code CLI (
~/.claude.json)
Manual Configuration:
If you prefer manual setup, add Blitz to your IDE's mcpServers configuration:
{
"mcpServers": {
"blitz": {
"command": "/Applications/Blitz.app/Contents/MacOS/Blitz",
"args": ["mcp"]
}
}
}
Exposed MCP Tools:
blitz_list_collections: Discover all collections, folders, and requests with folder breadcrumb paths (path,folder).blitz_get_request: Retrieve full request details (URL, headers, query params, body, auth).blitz_send_request: Execute a saved request through Blitz pipeline (returns parsedjson, supportssave_to_envauto-extraction, records history).blitz_execute_http: Run an ad-hoc HTTP request through Blitz engine (returns parsedjson, supportssave_to_envauto-extraction, records history).blitz_list_environments: List available environments, active state, and variable keys/values.blitz_set_env_var: Update or create an environment variable (e.g. auth token or baseUrl) directly from agent output.blitz_set_active_environment: Switch or clear the active environment.blitz_upsert_request: Create or edit request drafts in collections and folders.blitz_import_curl: Import raw cURL commands directly into collections.blitz_generate_code: Generate client code snippets (cURL,Python,TypeScript,Rust,Go,Java,C#) with variable interpolation.blitz_list_history: Query past request executions with status, latency, and timestamps.blitz_run_collection: Run a complete collection test suite (supportsfolderscoping) and return assertion results.blitz_suggest_assertions: Automatically generate test scripts & assertions (status, JSON path, headers, latency ceiling) from an HTTP response or request, with optional auto-attachment.blitz_perf_test: Run concurrent load/benchmark testing against any request or collection (concurrency VUs, duration) and get RPS throughput, error rate, and p50/min/max latencies.blitz_sync_openapi: Sync or import an OpenAPI 3.0 specification (JSON/YAML or file path) non-destructively, preserving custom tests, scripts, and auth.blitz_get_collection: Retrieve the full collection configuration tree (folders, requests, scripts, variables, auth) in a single call.blitz_create_environment: Create a new named environment with initial key-value variables and optional instant activation.blitz_delete_request: Delete an existing request from a collection by ID or name.blitz_start_mock: Spin up a local HTTP mock server (127.0.0.1:<port>) serving saved request examples or routes.blitz_stop_mock: Stop the active local mock HTTP server.blitz_export_collection: Export any collection as Postman v2.1 JSON or OpenAPI 3.0 specification.blitz_diff_responses: Compare two response payloads or compare live execution across two environments (stagingvslocal).blitz_delete_collection: Delete a collection by ID or name.blitz_generate_docs: Generate standalone HTML documentation (static single-file or Swagger UI) from collections.blitz_create_folder: Create nested organizational folders inside collections.blitz_delete_folder: Delete a folder and all contained requests recursively.blitz_delete_environment: Delete a named environment.blitz_clear_history: Clear execution history globally or filtered by request ID.blitz_import_har: Import HTTP Archive (.harfile path or raw JSON) into collections.blitz_manage_cookies: Inspect or clear session cookies stored in Blitz's global cookie jar.blitz_websocket_send: Connect to a WebSocket endpoint (ws://orwss://), transmit text/JSON payload, capture streaming inbound frames, and gracefully disconnect.blitz_grpc_unary: Execute a gRPC unary procedure call over HTTP/2 with JSON body encoding (google.protobuf.Struct) returning response data and status code.blitz_sse_listen: Connect to Server-Sent Events (SSE) HTTP stream, capture streaming events (event,data,id), and return trace.blitz_mqtt_publish: Publish a message to an MQTT broker topic with QoS 1 confirmation and disconnect.blitz_mqtt_subscribe: Subscribe to an MQTT topic on a broker and wait for incoming message with timeout.blitz_graphql_ws_subscribe: Execute GraphQL subscription over WebSocket (graphql-transport-ws) and collect streaming payloads.blitz_socketio_emit: Connect to Socket.IO server, emit event with payload, and capture inbound events.
Exposed MCP Resources:
blitz://collections: Hierarchy of all collections, folders, and requests.blitz://environments: Configured environments and variables.blitz://active-env: Currently active environment with resolved key-value pairs.blitz://history: Recent execution history.
Team Collaboration via Git
Sync team workspaces directly through Git without cloud intermediaries:
- Link Git Remote: In the top toolbar, select Team → Link Git Remote…
- Work Locally: Add endpoints, edit environments, and write tests.
- Sync: Click Sync — Blitz pulls changes, handles merges, and pushes through standard Git (SSH or
ghCLI). - Leak Prevention: Passwords, tokens, and local secret variables are kept local and excluded from commits.
License
Distributed under the MIT / Apache 2.0 open-source licenses. Not affiliated with or endorsed by Postman Inc.
Blitz MCP questions
What tools does Blitz MCP provide?
It exposes 37 tools, including blitz_list_collections, blitz_get_request, blitz_send_request, blitz_execute_http, blitz_list_environments.
What can AI agents do with the Blitz MCP server?
AI agents like Cursor and Claude can list collections, inspect request schemas, execute endpoints, and run API test suites locally.
Does Blitz require a cloud account?
No, Blitz is 100% offline and local-first, meaning no accounts or cloud tracking are required.
How does Blitz handle sensitive credentials?
It uses native OS Keychain integration to ensure secrets stay encrypted locally.