Use this server only for APKs and Android applications you own, are responsible for, or are explicitly authorized to inspect., The plugin exposes decompiled classes, methods, fields, smali, manifests, strings, resources, xrefs, and debugger state to the MCP client., Rename and refactor tools can modify JADX project state and naming decisions; review changes before saving project output or using generated reports., The architecture docs describe a local plugin service with no built-in authentication; keep it bound to localhost unless you add network controls., Optional HTTP mode for the MCP server should not be exposed to untrusted networks without authentication, TLS, and firewall restrictions., Decompiled strings, manifests, and resources can contain prompt-injection text or untrusted content; treat tool output as untrusted input.
Privacy notes
APK source, package names, manifests, resources, strings, selected text, debugger variables, and analysis prompts may be sent to the model provider., Debugger tools can expose runtime values, tokens, identifiers, device data, or user information from the analyzed application., Reverse-engineering work can reveal proprietary code, licensed assets, customer data, or confidential security findings., Avoid uploading malware samples, third-party apps, client applications, or regulated data to external model providers without approval.
Author
zinja-coder
Submitted by
oktofeesh1
Claim status
unclaimed
Last verified
2026-06-05
Decision playbook
Review trust signals before you adopt
Signals are present but mixed. Use the checklist below to confirm the source and operational safety for your environment.
Compare context
Selected
0
Current score
63
Baseline
—
Delta
No baseline selected
No major trust-signal divergence detected in the current selection.
Source and provenance checks
Needs review
Confirm ownership and provenance before trusting install instructions.
Source link availableRequired
Open the canonical repository and verify ownership.
Done
Source provenance statusRequired
Marked as source-backed.
Done
Metadata reviewed
No reviewed flag detected in metadata.
Pending
Safety and privacy checks
Complete
Validate risk disclosures before installation or API wiring.
Safety notes presentRequired
Review the listed safety guidance before running commands.
Done
Privacy notes presentRequired
Review data handling notes before connecting accounts or secrets.
Done
Trust level risk gateRequired
Trust level does not block evaluation.
Done
Package and install checks
Needs review
Check package metadata and artifact integrity signals.
Install payload available
Install or copy payload is available for review.
Done
Package verification flag
No package verification flag provided.
Pending
Checksum metadata
No checksum provided for downloaded artifact.
Pending
Compare-driven decision checks
Needs review
Use compare context to validate trade-offs before adoption.
Compare tray has multiple entries
Add at least one more entry to compare trust differences.
6 safety and 4 privacy notes across 5 risk areas. Review closely: credentials & tokens, permissions & scopes, network access, third-party handling.
5 areas
SafetyPermissions & scopesUse this server only for APKs and Android applications you own, are responsible for, or are explicitly authorized to inspect.
SafetyGeneralThe plugin exposes decompiled classes, methods, fields, smali, manifests, strings, resources, xrefs, and debugger state to the MCP client.
SafetyGeneralRename and refactor tools can modify JADX project state and naming decisions; review changes before saving project output or using generated reports.
SafetyNetwork accessThe architecture docs describe a local plugin service with no built-in authentication; keep it bound to localhost unless you add network controls.
SafetyNetwork accessOptional HTTP mode for the MCP server should not be exposed to untrusted networks without authentication, TLS, and firewall restrictions.
SafetyGeneralDecompiled strings, manifests, and resources can contain prompt-injection text or untrusted content; treat tool output as untrusted input.
PrivacyThird-party handlingAPK source, package names, manifests, resources, strings, selected text, debugger variables, and analysis prompts may be sent to the model provider.
PrivacyCredentials & tokensDebugger tools can expose runtime values, tokens, identifiers, device data, or user information from the analyzed application.
PrivacyGeneralReverse-engineering work can reveal proprietary code, licensed assets, customer data, or confidential security findings.
PrivacyNetwork accessAvoid uploading malware samples, third-party apps, client applications, or regulated data to external model providers without approval.
Disclosure: Community-maintained reverse-engineering tooling for JADX; review the live repositories, releases, and security policy before using it in sensitive work.
Safety notes
Use this server only for APKs and Android applications you own, are responsible for, or are explicitly authorized to inspect.
The plugin exposes decompiled classes, methods, fields, smali, manifests, strings, resources, xrefs, and debugger state to the MCP client.
Rename and refactor tools can modify JADX project state and naming decisions; review changes before saving project output or using generated reports.
The architecture docs describe a local plugin service with no built-in authentication; keep it bound to localhost unless you add network controls.
Optional HTTP mode for the MCP server should not be exposed to untrusted networks without authentication, TLS, and firewall restrictions.
Decompiled strings, manifests, and resources can contain prompt-injection text or untrusted content; treat tool output as untrusted input.
Privacy notes
APK source, package names, manifests, resources, strings, selected text, debugger variables, and analysis prompts may be sent to the model provider.
Debugger tools can expose runtime values, tokens, identifiers, device data, or user information from the analyzed application.
Reverse-engineering work can reveal proprietary code, licensed assets, customer data, or confidential security findings.
Avoid uploading malware samples, third-party apps, client applications, or regulated data to external model providers without approval.
Prerequisites
Java 11 or newer and JADX-GUI installed.
Python and uv available for installing or running the companion MCP server.
The JADX AI MCP plugin installed in JADX-GUI.
An Android APK, JADX project, or reverse-engineering target you are authorized to analyze.
Permission to expose decompiled code, resources, manifests, and debugger state to the configured MCP client and model provider.
Schema details
Install type
cli
Troubleshooting
No
Source repository stats
Scope
Source repo
Collection metadata
Estimated setup
20 minutes
Difficulty
advanced
Tool listing metadata
Disclosure
Community-maintained reverse-engineering tooling for JADX; review the live repositories, releases, and security policy before using it in sensitive work.
JADX AI MCP combines a Java plugin for JADX-GUI with a companion Python MCP
server. Together they let Claude and other MCP clients inspect the currently
loaded Android APK, query decompiled classes, search methods, inspect resources,
retrieve manifest data, follow references, and interact with debugger state.
The architecture is a local bridge: the MCP client talks to the Python server,
and the Python server forwards tool calls to the JADX-GUI plugin. The upstream
docs describe stdio mode for normal MCP clients and optional HTTP mode when a
separate transport is needed.
These sources were reviewed on 2026-06-05. Prefer the live plugin repo,
companion server repo, ReadTheDocs site, installation guide, architecture guide,
server source, project metadata, licenses, and security policy for current
requirements, transport flags, tool names, release packaging, and vulnerability
reporting guidance.
Features
Fetch the current class, selected text, class source, smali, fields, methods, and package tree from JADX-GUI.
Search classes and methods across a decompiled APK with pagination.
Retrieve AndroidManifest.xml content, manifest components, strings, and resource files.
Inspect main activity and application classes derived from manifest package data.
Follow references to classes, methods, and fields.
Read debugger stack frames, threads, and variables from JADX debugger context.
Rename classes, methods, fields, packages, and variables through JADX refactoring tools.
Run through a stdio MCP server, with optional HTTP mode for advanced local transport setups.
Installation
Install the JADX plugin using the upstream plugin command:
Then install or run the companion Python MCP server following the upstream
installation guide. The simplified MCP client configuration uses the installed
jadx_mcp_server command:
Open an APK in JADX-GUI before invoking tools so the plugin has a loaded project
to inspect. Keep the plugin and MCP server local unless you have reviewed the
transport flags, binding addresses, and network controls in the documentation.
Use Cases
Ask Claude to summarize an Android class, package, or selected decompiled method.
Search for suspicious strings, permissions, exported components, or API usage patterns in an authorized APK.
Retrieve manifest components and resources before writing an analysis report.
Inspect smali and decompiled Java side by side while reversing obfuscated code.
Follow cross-references to understand how a class, method, or field is used.
Use debugger context to explain stack frames, thread state, and runtime variables.
Rename obfuscated identifiers inside JADX project state before continuing manual review.
Safety and Privacy
JADX AI MCP is powerful reverse-engineering tooling, so keep authorization and
data boundaries explicit. Do not use it on apps, malware samples, customer APKs,
or third-party code unless you have permission and know where model context will
be stored.
The architecture docs state that the JADX plugin binds locally and has no
built-in authentication. Leave services on localhost by default, and add
authentication, TLS, firewall rules, and host isolation before considering any
remote HTTP setup. Treat decompiled code, strings, resources, and debugger
values as sensitive and untrusted because they can contain proprietary logic,
secrets, personal data, or prompt-injection content.
Duplicate Check
Existing reverse-engineering entries cover other IDA and Ghidra MCP servers,
but no zinja-coder/jadx-ai-mcp, zinja-coder/jadx-mcp-server, JADX AI MCP
entry, or matching source URL was found in content/mcp.
Show that JADX AI MCP Server is listed on HeyClaude. Paste this Markdown into your README — it renders the badge and links back to this page.
[](https://heyclau.de/entry/mcp/jadx-ai-mcp-server)
How it compares
JADX AI MCP Server side by side with 3 alternatives on trust, install, platform support, and disclosed safety notes — all from reviewed registry metadata.
Apache-licensed Ghidra MCP bridge with GUI and headless workflows for binary analysis, decompilation, scripting, debugger access, batch operations, and Ghidra Server collaboration.
Ghidra plugin and MCP bridge that lets AI assistants inspect, decompile, rename, comment, and analyze binaries through Ghidra reverse-engineering workflows.
MCP server and Claude Code plugin for connecting IDA Pro or headless idalib to AI assistants for reverse engineering, decompilation, xref lookup, renaming, commenting, and binary analysis workflows.
✓Use this server only for APKs and Android applications you own, are responsible for, or are explicitly authorized to inspect.
The plugin exposes decompiled classes, methods, fields, smali, manifests, strings, resources, xrefs, and debugger state to the MCP client.
Rename and refactor tools can modify JADX project state and naming decisions; review changes before saving project output or using generated reports.
The architecture docs describe a local plugin service with no built-in authentication; keep it bound to localhost unless you add network controls.
Optional HTTP mode for the MCP server should not be exposed to untrusted networks without authentication, TLS, and firewall restrictions.
Decompiled strings, manifests, and resources can contain prompt-injection text or untrusted content; treat tool output as untrusted input.
✓Use this server only for binaries and systems you are authorized to inspect, debug, emulate, modify, document, or reverse engineer.
The tool surface includes write-capable operations such as renaming, typing, commenting, label changes, script execution, structure creation, and project/version-control workflows.
Unknown binaries, malware samples, debugger targets, and imported scripts can be dangerous; isolate analysis environments and avoid running untrusted code on a primary workstation.
Headless, Docker, CI, and Ghidra Server workflows can affect shared repositories or automated pipelines if configured with write access.
Keep human review on operations that change analysis state, scripts, project metadata, or shared Ghidra Server content.
✓Reverse engineering may be legally restricted; confirm authorization before analyzing third-party or proprietary binaries.
Malware, exploit samples, and unknown binaries should be handled in isolated environments with limited network and filesystem access.
LLM-generated names, comments, and conclusions must be verified against Ghidra decompiler output, disassembly, imports, exports, and xrefs.
Keep backups of Ghidra projects before allowing an agent to rename symbols, alter comments, or save project state.
✓Reverse engineering may be legally or contractually restricted; confirm authorization before analyzing proprietary binaries or third-party samples.
Malware samples and exploit tooling should be handled in isolated environments with network and filesystem controls.
LLMs can hallucinate reverse-engineering conclusions; verify findings against disassembly, decompiler output, xrefs, and reproducible scripts.
Do not let an agent patch, rename, comment, or save IDA databases used for evidence without review and backups.
Privacy notes
✓APK source, package names, manifests, resources, strings, selected text, debugger variables, and analysis prompts may be sent to the model provider.
Debugger tools can expose runtime values, tokens, identifiers, device data, or user information from the analyzed application.
Reverse-engineering work can reveal proprietary code, licensed assets, customer data, or confidential security findings.
Avoid uploading malware samples, third-party apps, client applications, or regulated data to external model providers without approval.
✓Tool calls may expose binary names, file paths, hashes, strings, imports, exports, symbols, decompiled code, comments, type names, project names, repository URLs, debugger state, memory reads, and trace metadata.
Reverse engineering projects often contain proprietary firmware, customer binaries, vulnerability findings, credentials embedded in samples, exploit indicators, and private research notes.
MCP client logs, model transcripts, Ghidra project history, Docker volumes, CI logs, and Ghidra Server repositories may retain sensitive analysis data.
Do not commit private binaries, generated Ghidra projects, debugger traces, analysis exports, credentials, proprietary symbols, or customer-specific reverse engineering notes.
✓Binaries, function names, strings, imports, exports, decompilation output, comments, and analysis notes may be sent to the MCP client and model.
Extracted strings and decompiled code can expose proprietary logic, credentials, API endpoints, malware indicators, or customer data.
Generated reports can reveal sensitive vulnerability research or product internals.
✓Binaries, IDA databases, symbol names, strings, comments, decompiler output, and vulnerability findings may be sent to the MCP client and model.
Proprietary firmware, customer crash samples, malware indicators, license keys, and embedded secrets can appear in extracted strings or decompiled code.
Reports and prompts may reveal unreleased product internals or sensitive security research.
Prerequisites
Java 11 or newer and JADX-GUI installed.
Python and uv available for installing or running the companion MCP server.
The JADX AI MCP plugin installed in JADX-GUI.
An Android APK, JADX project, or reverse-engineering target you are authorized to analyze.
Ghidra 12.1 or a compatible version installed and available on the local machine or analysis host.
Java 21, Maven 3.9 or newer, Python 3.10 or newer, and Python package installation available.
Ghidra extension deployment permissions for the user profile or headless environment.
Authorization to analyze the binaries, firmware, traces, symbols, and shared Ghidra projects loaded into the tool.
Installed Ghidra.
Python 3.
MCP Python SDK dependencies required by the project.
Latest GhidraMCP release zip imported as a Ghidra extension.
IDA Pro 8.3 or newer, with IDA Pro 9 recommended; IDA Free is not supported.
Python 3.11 or newer and uv.
Globally activated idalib for the headless Claude Code plugin path.
Supported MCP client such as Claude Code, Claude Desktop, Codex, Cursor, Gemini CLI, or another documented host.