| Name | Modified | Size | Downloads / Week |
|---|---|---|---|
| Parent folder | |||
| neurosploit-v4.3.1-windows-x64.zip | 2026-10-15 | 6.4 MB | |
| neurosploit-v4.3.1-macos-arm64.tar.gz | 2026-10-15 | 6.2 MB | |
| neurosploit-v4.3.1-macos-x64.tar.gz | 2026-10-15 | 6.5 MB | |
| neurosploit-v4.3.1-linux-x64.tar.gz | 2026-10-15 | 6.8 MB | |
| neurosploit-v4.3.1-linux-arm64.tar.gz | 2026-10-15 | 6.7 MB | |
| README.md | 2026-10-15 | 1.5 kB | |
| v4.3.1 source code.tar.gz | 2026-10-15 | 1.4 MB | |
| v4.3.1 source code.zip | 2026-10-15 | 1.8 MB | |
| Totals: 8 Items | 35.8 MB | 0 | |
NeuroSploit v4.3.1
Fixes and new capabilities since v4.3.0.
Fixes
- Subscription auth: the Claude/Codex CLI backends no longer fail with "Invalid API key" when a stray
ANTHROPIC_API_KEY/OPENAI_API_KEYis in the environment — the subscription login is always used on the CLI path. - Scope: an endpoint's HTTP method ("GET /api/x") is no longer mistaken for its host, so host-relative findings are no longer wrongly quarantined as out-of-scope and the attack graph no longer grows bogus
HOST "get"/"post"nodes. - Web console: reopening a running run resolves the run-id to the live job (activity log streams again); the pinned-leads header collapses instead of burying the tabs; live findings and per-agent progress are parsed from the decorated REPL output.
New
- GLM 5.1 / 5.2 / 5.3 — via the Z.ai API, and agentically through the Claude Code CLI using Z.ai's Anthropic-compatible endpoint (the Z.ai devpack / cline / zcode setup) with
--subscription. claude-opus-4-6added to the model catalog.- Import an API collection (Postman / OpenAPI / Swagger / HAR, JSON) as known test surface —
--collection <file>(CLI),/collection <file>(REPL), or the Creds panel in the web console.
Install
:::bash
curl -fsSL https://raw.githubusercontent.com/JoasASantos/NeuroSploit/main/setup.sh | bash
Prebuilt asset: macOS arm64 (binary + agent library). Other platforms build from source (the installer handles it).
