Visually design, automate, and diagnose complex workflows across protocols with a built-in Agentic AI Copilot. 7 core protocols ship in the binary; extend with one-click plugins. A cross-platform desktop app with a powerful CLI — all in one binary.
Current version: loading...
Visual debugging, AI tool evaluation, and automated end-to-end multi-protocol testing in one workspace.
Orchestrate multi-protocol DAGs from natural language prompts. Features autonomous goal execution, midway steering, 1-click execution error diagnosis and smart repair, and local codebase awareness.
Debug Model Context Protocol (MCP) servers via stdio or SSE. Auto-discover tools and prompts, chain invocations, and evaluate output accuracy with visual assertions.
Connect HTTP triggers, Kafka queues, Redis caches, and SQL databases across a single canvas. Eliminate fragmented Python/Go test scripts.
Native JUnit XML reports (--report-format junit --report-file results.xml) plug straight into Jenkins, GitHub Actions, and GitLab CI — every node maps to a test case, so PRs surface flow results without custom tooling.
Decouple environment sets (dev, staging, prod) into git-shared YAML files, paired with git-ignored .local.yaml private overrides for credentials. Features 4-tier cascading resolution, credential leak scanners, and headless CI management via mpe env.
Core protocols run out of the box; everything else installs with one command
GraphQL
Why modern engineering and QA teams choose MPE Flow over single-protocol tools
| Core Dimension | MPE Flow | Postman | Apache JMeter | Bruno |
|---|---|---|---|---|
| Multi-Protocol Chaining | ✓ Native HTTP, WS, TCP, UDP, SSE, GQL, Kafka, MCP | HTTP only (isolated WS tab) | Supported via complex sampler plugins | HTTP & basic GraphQL only |
| Engine & Resource Footprint | ⚡ Native Rust (Zero GC, <30MB RAM) | Electron (Heavy memory usage) | Java JVM (High RAM & GC pauses) | Electron / Node.js |
| Delivery & Mode | ✓ Single Binary (Desktop GUI + CLI) | Desktop App + Newman wrapper | Java GUI + Shell script | Desktop App + Bru CLI |
| MCP (Model Context Protocol) Eval | ✓ Native Visual Inspector & Assertions | ✗ Not Supported | ✗ Not Supported | ✗ Not Supported |
| Built-in Stress Testing | ✓ Native Tokio Async (p50/p90/p99 latency) | ✗ Paid cloud monitor only | ✓ Built-in (Thread-heavy) | ✗ Not Supported |
| Local-First & Data Sovereign | ✓ 100% Local Git-friendly .mpf files | ✗ Mandatory Cloud Account | ✓ Local XML/JMX | ✓ Local plain files |
| Built-in AI Copilot & Auto-Fix | ✓ Full-canvas Copilot (NL-to-DAG, auto-layout, autonomous run, 1-click error fix, code perception) | Postbot (HTTP-only, cloud account required, paid tier) | ✗ Not Supported | ✗ Not Supported |
| CI/CD Native Reporting | ✓ JUnit XML, JSON, HTML, Markdown | Newman CLI / JSON | JTL / HTML report | CLI text output |
Everything you need to know about MPE Flow architecture, protocols, and deployment.
The AI Copilot connects to any OpenAI-compatible provider (such as DeepSeek, OpenAI, Claude, or local Ollama instances) and interacts directly with the MPE Flow canvas via tool calling. It can create and connect nodes, compute adaptive layouts, write JavaScript/Rhai scripts, inspect local codebases, execute workflows autonomously to reach a goal, and diagnose/repair execution errors in a single click — with zero data lock-in and offline model support.
MPE Flow (Multi-Protocol Flow Executor) is a high-performance visual workflow orchestration and integration testing engine built with Rust and Tauri. It enables engineers to chain and automate complex scenarios across HTTP, WebSocket, TCP, UDP, SSE, GraphQL, Kafka, and MCP in a single desktop application and standalone CLI.
MPE Flow includes 7 native core protocols inside the single binary: HTTP/HTTPS, WebSocket, TCP, UDP, SSE (Server-Sent Events), GraphQL, and Script/Assertion nodes. Additional protocols such as Kafka, MCP, gRPC, Redis, MongoDB, SMTP, and IMAP are available through one-click sidecar plugins.
Unlike HTTP-centric tools like Postman, MPE Flow provides native first-class support for stateful protocols (TCP, WebSocket, SSE streams, Kafka pub/sub, MCP) with visual graph-based flow design, branching, variable extraction, assertion chains, and high-concurrency stress testing in a lightweight Rust binary.
Yes. MPE Flow ships as a dual-mode single binary containing both the GUI and the standalone mpe CLI. You can run mpe run flow.mpf -e staging --report-format junit --report-file results.xml directly in GitHub Actions, GitLab CI, Jenkins, or Docker without GUI dependencies.
Yes, MPE Flow is open source under the MIT License, hosted on GitHub at github.com/multi-protocol-flow.
Join developers who trust MPE Flow for their integration testing and automation needs.
