Reverser Space 101

Give your agent
somewhere real
to work.

Reverser Space is the collaboration and execution layer for reverse engineering. Humans and AI agents work on the same binary, with the same tools, context, and live state.

Enter a live session → See how it works

Real binary. Read-only. No account needed.

firecracker-test-sample LIVE DEBUGGER
The real Reverser Space debugger showing dynamic disassembly, decompiled code, registers, and a stopped process.
Real platform capture Dynamic listing, decompiler, registers, stack, memory, breakpoints, and terminal share one run.
One session DisassemblyDecompilerDebuggerMemoryGraphsNotesMCP

The simple version

A runtime,
not a wrapper.

Most AI integrations give a model a chat box or a few commands. Reverser Space gives it the whole lab, then puts the expert in the same room.

The usual flow

Ask. Wait. Read a report.

  • Context copied into a chat
  • Analysis hidden behind an answer
  • Corrections lost in another thread
  • One tool on one machine

The Reverser Space flow

Watch. Guide. Keep working.

  • Humans and agents share one binary
  • Every action appears live
  • Findings stay attached to the analysis
  • Experts can correct or take over

One shared unit of work

Everything happens inside a session.

A session is one binary plus its durable analysis, notes, findings, chat, participants, and activity history. Choose a layer to see what it contributes.

01 / Binary

The evidence stays central.

Every claim can lead back to the exact function, address, byte range, or runtime state behind it.

Built for the whole crew

Choose your seat.

For reverse engineers

Work with an agent without giving up the wheel.

Delegate the first pass, repetitive navigation, or a focused research question. Watch every step, correct the analysis in place, and keep the useful work.

OutcomeMore ground covered, with expert judgment still in control.

Real security workflows

Pick a mission.

The lab stays the same. The question, tools, and balance between human and agent change with the work.

MISSION BRIEFTRIAGE

01 / TRIAGE

Turn an unknown binary into a map.

Let an agent rank interesting functions, identify imports and behaviors, and form initial hypotheses. An analyst validates the evidence and chooses where to go deep.

Rank functionsInspect signalsWrite hypothesesChoose targets

Use Reverser Space only for binaries and systems you own or are explicitly authorized to test.

The workspace is the proof

Every claim leads back to code.

Open the function behind a finding. Compare decompiled C with disassembly. Follow a string to its references. Move from static analysis to runtime state without losing the thread.

The result is not a report about the analysis. It is the analysis, still alive and ready for the next question.

The real Reverser Space analysis workspace with sections, functions, symbols, static listing, Ghidra decompilation, notes, activity, and chat. Shared state Live presence Real decompiler

Your first five minutes

See it. Try it.
Bring your agent.

Check off the path as you go. Progress stays in this browser.

The real Reverser Space session picker with an upload target, activity history, and a shared live session.
Start here Drop a binary, attach a matching PDB if you have one, and enter the session when analysis is ready.
0/ 4 complete
01 Watch a live sessionOpen Notes, click a function, and follow the evidence into code. Open demo →
02 Start your own sessionUpload a binary, optionally attach a matching PDB, then analyze. Open app →
03 Leave the analysis betterRename one function, follow one xref, and record one conclusion. Press ? for shortcuts
04 Add a collaboratorInvite a teammate or agent, or connect your own MCP client. MCP guide →

Model agnostic by design

One MCP server.
Every session.

Connect Claude Code, Cursor, Claude Desktop, or any HTTP MCP client. Your agent discovers the sessions you can access and uses your role on each one.

Connect your agent →
Account endpointOAuth or bearer token
https://api.reverser.space/mcp/account

One connection routes tools with an explicit session ID.

✓ Your model✓ Your permissions✓ Attributed actions

Trust is part of the runtime

Operational does not
mean opaque.

01

Attributed

Every meaningful change shows who made it and which human authority stands behind an agent.

02

Permission-capped

Agents inherit the user's session role. Access is checked again before every operation.

03

Correctable

Experts can inspect, revise, redirect, take over the debugger, or stop a run.

04

Persistent

Useful work stays in the saved analysis instead of disappearing into a transcript.

Share linkOne session, read-only
ViewerInspect and follow
EditorAnalyze and change
CreatorOwn and manage
AgentInherits human access

Cloud by default

Your lab, ready
when the link opens.

The standard experience is hosted. Analysis runs server-side, collaborators enter through a URL, and nothing needs to be installed on each analyst's workstation.

Available on request

Dedicated managed tenant

Organizations with stronger isolation, compliance, or data-boundary requirements can request a managed single-tenant deployment of the complete stack.

Discuss your environment →

Live debugging currently focuses on native Linux x86-64 ELF samples. Agent and debugger availability can vary by deployment configuration.

Never Ride Solo

Give your agent
somewhere to work.

Open a real analysis session, watch humans and agents share the work, then bring your own binary or your own agent.

Watch an agent work → Open the app Book a demo