Skip to content

Latest commit

 

History

History
65 lines (55 loc) · 2.96 KB

stack.md

File metadata and controls

65 lines (55 loc) · 2.96 KB

The stack

rekit is layered so every layer can be used and verified on its own. Nothing executes a sample: each backend is a read-only parser or a read-only analysis binary.

                     +-------------------------------+
  CLI / scripts ---> |  rekit.py  (13 subcommands)   |
                     |  triage strings iocs disasm   |
                     |  apk firmware js pcap ghidra  |
                     |  capa floss yara r2           |
                     +---------------+---------------+                              
                                     |
                                     v
                     +-------------------------------+
                     |  read-only backends           |
                     |  readelf objdump radare2 jadx |
                     |  binwalk tshark YARA capa     |
                     |  FLOSS Ghidra headless        |
                     +-------------------------------+
                                     ^
                                     |
                     +-------------------------------+
  MCP clients  ----> |  mcp_server.py (13 tools re_*)|
  (stdio / HTTP)     |  FastMCP "re-connector"       |
                     +-------------------------------+
                                     ^
                                     |
                     +-------------------------------+
  agents       ----> |  REA (rea-agents, optional)   |
                     |  125 MCP tools / 80 CLI cmds  |
                     |  evidence envelopes (sha256)  |
                     +-------------------------------+

Layer 1 - rekit CLI

Single file, no service, no state. Every subcommand prints JSON on --json and reports a missing backend as tool_missing instead of guessing. This is the layer to script and to put in CI.

Layer 2 - rekit MCP server

mcp_server.py exposes the same work as 13 MCP tools (re_triage, re_strings, re_iocs, re_disasm, re_apk, re_firmware, re_js, re_pcap, re_ghidra_decompile, re_capa, re_floss, re_yara, re_health) over stdio or HTTP. re_health probes all 18 backends and reports which are installed, so a client can tell "clean file" from "no tool available" before trusting a result.

Layer 3 - REA (optional)

REA is an independent agent-directed layer that drives Ghidra/IDA/Hopper and returns sha256-pinned evidence envelopes (confidence, authority, limitations) rather than bare text. It is optional: rekit works without it. When present, its 125 tools sit on the same Ghidra install that rekit ghidra uses, so both layers see identical binaries - see docs/verification.md for the cross-surface hash check.

Choosing a layer

Need Use
One-off triage in a shell / CI rekit.py (CLI)
Let an LLM agent reason over a binary mcp_server.py (13 tools)
Deep, multi-step analysis with evidence trail REA (125 tools)
Anything that must not run the sample all three - static only