MIT-licensed reverse-engineering toolkit (CLI + MCP) for AI agents; connects to DeepSeek Harness as a standard MCP server — the README does not mention DSH, so integration is via generic MCP registration.
DSH integration
Compatible
Author-claimed
Safety audit
Unaudited
Last verified
2026-08-21
License
MIT
01What can it help you accomplish?
Investigate how an application feature works — from app behavior down to native binaries — using an agent
Agent explanations of feature internals backed by pseudocode, strings, cross-references and call graphs, recorded as reproducible Evidence v2 records
Developers, security and interop engineers who want a coding agent to understand closed-source or undocumented software
Recreate an investigated feature in your own project during the same coding session
A version of the feature adapted to your stack, interface, and requirements, built by the agent with its normal file-editing tools
Product developers porting or reimplementing a feature whose source code they do not have
Analyze proprietary binaries locally without sending the app to a hosted service
Local decompilation, artifact inventory and observation results; evidence and snapshots stay on the local host
Teams handling proprietary or sensitive binaries that cannot leave the machine
02How to install into DeepSeek Harness
Prerequisites
- Node.js 22.19+ or 24.11+ (including newer releases) and npm
- macOS 12 or newer, Ubuntu 24.04+, Fedora 41+, or 64-bit Arch Linux (Windows x64 only for the experimental, Ghidra-only native PE P0 boundary)
- For deep binary analysis: Hopper (separate desktop application with its own license) or a bring-your-own Ghidra 12.1.2 installation with 64-bit JDK 21 on Linux x64
- An MCP-capable agent (e.g. DeepSeek Harness) — setup auto-detects Claude Code, Claude Desktop, Codex, Cursor, Gemini CLI and Windsurf; other agents use manual MCP configuration
Installation steps
- 01
Install the CLI globally: `npm install --global rea-agents` (or run without installing via `npx --yes rea-agents@latest setup`)
$ npm install --global rea-agents
- 02
Run `rea setup` (or `npx --yes rea-agents@latest setup`), review the printed plan, and choose Agent Integration — REA installs the pinned MCP registration and its matching routing skill as one transaction
$ npx --yes rea-agents@latest setup
- 03
Restart the configured agent so it loads REA
- 04
For agents not auto-detected (e.g. DeepSeek Harness), add the manual MCP config: `{ "mcpServers": { "rea": { "command": "npx", "args": ["-y", "rea-agents@3.1.0", "mcp"] } } }`
Verify the integration
- Run `rea doctor` (or `npx -y rea-agents@latest doctor`): a read-only diagnostic that distinguishes unsupported hosts, missing dependencies, a missing local analysis engine, configuration drift, and healthy checks
- After a successful setup, REA reports the capabilities now ready to use and a concrete next step, and only claims an integration is ready when setup's final diagnostic check verified it
Rollback
- `rea uninstall` removes only REA-owned MCP registrations and the managed skill (`--purge-data` also removes ~/.rea/cache and ~/.rea/state); Hopper, Node.js, evidence, unrelated skills and other MCP servers are preserved
03DSH integration and capability boundaries
Standard MCP server registered via manual mcpServers config (`npx -y rea-agents@<version> mcp`); the README does not mention DeepSeek Harness, so integration is generic MCP rather than an officially documented DSH plugin.
Native binary analysis via Hopper / Ghidra
Mach-O, ELF, PE, .app, ZIP, APK, IPA, ASAR, plist, JavaScript and source-map targets→Pseudocode, assembly, strings, names, callers/callees, xrefs, call graphs and bounded function dossiers
Launches and controls Hopper (or a caller-selected Ghidra bridge) locally; Hopper may come to the foregroundWrites Evidence v2 records and optional analysis snapshots to operator-approved local rootsJavaScript / Electron / browser observation
Local JavaScript/Electron app directories or ASAR archives; a user-owned Chrome-family browser on a loopback CDP endpoint→Static application mapping without execution, passive CDP page capture, and bounded source-map reconstruction
Browser and V8 Inspector observation are disabled by default and require explicit environment-variable approvalPlaywright scenario capture is a separately approved, explicitly mutating boundaryReproducible Evidence v2 records
Every successful analysis result→Deterministic Evidence v2 bundles with artifact and provider identity, confidence, authority, limitations and locations; import/export/compare across sessions
Writes evidence bundles only to operator-approved absolute roots; imports never execute bundle contentSame workflows over CLI and MCP
Any agent that supports local MCP servers, or a terminal→Identical reverse-engineering workflows from CLI and MCP, six guided MCP prompt workflows, and setup that auto-configures Claude Code, Claude Desktop, Codex, Cursor, Gemini CLI and Windsurf
`rea setup` writes additive, backup-first MCP registrations into detected agent configuration files
04Who is it for? When not to use it?
Good for
- Developers, security and interop engineers who want a coding agent to understand closed-source or undocumented software
- Product developers porting or reimplementing a feature whose source code they do not have
- Teams handling proprietary or sensitive binaries that cannot leave the machine
Not for
- Supported hosts are macOS 12+, Ubuntu 24.04+, Fedora 41+ or 64-bit Arch Linux; Windows x64 is limited to the experimental, Ghidra-only native PE P0 boundary, and some current workflows still require Hopper and macOS.
- Deep binary analysis requires a separate provider: Hopper (its own license, optionally installed by setup) or bring-your-own Ghidra 12.1.2 with JDK 21 — REA never downloads Ghidra or installs Java.
- REA's shipped providers, passive observers and Process Capture are not security sandboxes — providers and launched targets run with the current user's permissions; dynamic capabilities are disabled by default and require operator policy plus explicit per-call approval.
05Compatibility, maintenance and safety notes
- Supported hosts are macOS 12+, Ubuntu 24.04+, Fedora 41+ or 64-bit Arch Linux; Windows x64 is limited to the experimental, Ghidra-only native PE P0 boundary, and some current workflows still require Hopper and macOS.
- Deep binary analysis requires a separate provider: Hopper (its own license, optionally installed by setup) or bring-your-own Ghidra 12.1.2 with JDK 21 — REA never downloads Ghidra or installs Java.
- REA's shipped providers, passive observers and Process Capture are not security sandboxes — providers and launched targets run with the current user's permissions; dynamic capabilities are disabled by default and require operator policy plus explicit per-call approval.
MIT · actively maintained (latest release rea-agents-3.1.0, 2026-08-09)
06Frequently asked questions
How do I use REA inside DeepSeek Harness?
The README does not document a DeepSeek Harness-specific integration, but REA is a standard MCP server usable by any agent that supports local MCP servers. Run `npx --yes rea-agents@latest setup` and choose Agent Integration, or add the manual `mcpServers` entry pointing at `npx rea-agents mcp`, then restart the agent.
What are the prerequisites?
Node.js 22.19+ or 24.11+ with npm, and macOS 12+, Ubuntu 24.04+, Fedora 41+ or 64-bit Arch Linux (Windows x64 only for the experimental Ghidra-only PE boundary). Deep binary analysis additionally needs Hopper (separate license) or a bring-your-own Ghidra 12.1.2 + JDK 21 installation.
Does REA upload my app to the cloud?
No. REA has no hosted analysis service — current providers analyze artifacts and capture behavior locally. Your agent or model provider may have its own data policy, so review that separately.
Can REA recover the original source code?
No decompiler can guarantee the original source. REA gives the agent pseudocode, assembly, symbols, strings, metadata and relationships it can use to explain or compatibly recreate observed behavior.
How do I verify the installation?
Run `npx -y rea-agents@latest doctor` — a read-only diagnostic that distinguishes unsupported hosts, missing dependencies, a missing local analysis engine, configuration drift, and healthy checks.
07Related DSH workflows
ruflo
by ruvnet
🌊 The original agent meta-harness. Deploy intelligent multi-player swarms, coordinate autonomous workflows, and build conversational AI systems. Features adaptive memory, self-learning intelligence, RAG integration, and native Claude Code / Codex / Hermes and many more Integrated
reactive-resume
by amruthpillai
A one-of-a-kind resume builder that keeps your privacy in mind. Completely secure, customizable, portable, open-source and free forever. Try it out today!
everos
by evermind-ai
One portable memory layer for every AI agent: local-first, Markdown-native, user-owned, and self-evolving across apps, tools, and workflows.
yao
by yaoapp
✨ All your agents and workspaces in one place, on every device you own. Track tasks on a board, accessible from desktop, mobile, browser, or API. Self-hosted.
08Data and sources
Any agent that supports local MCP servers can use REA with the configuration below.
"args": ["-y", "rea-agents@3.1.0", "mcp"]
This page is generated from the project’s public documentation, repository metadata and a structured parse of DSH Plugins; last verified on 2026-08-21. Found an error? Submit a correction.
Best DeepSeek Harness Plugins
Twelve plugins worth installing first — picked from the whole catalog, across every category.