Bravos
Bravos is an npm CLI that turns the coding agent you already use (Claude Code, Codex, Cursor, opencode, omp or Antigravity) into a power user of four tools:
- guardlink, the threat model, written as external sidecars or inline annotations anchored to the code they describe.
- cxg, authorized pentests as code, against targets you own.
- The code graph, which answers routes, callers, reach and blast radius without grepping.
- SAST, a deterministic analyzer that reads the threat model and the graph.
bravos setup puts skills, session hooks and MCP wiring into a repository.
The agent in your session then does the work with the real tools. There is no
Bravos server and no Bravos database: the threat model lives in your
repository, the evidence comes from the tools, and your team shares it through
git.
Start from what your team already wrote
Section titled “Start from what your team already wrote”Give Bravos your PRDs, design documents and an earlier threat model. It keeps
them in the repository under .bravos/context/, drafts what the app is meant
to do for a person to confirm, and every coding agent that works on the
repository then judges code against that intent. See
context from your design documents.
The loop
Section titled “The loop”Every claim moves through the same steps:
- An annotation says code is exposed, or mitigated.
- It becomes a cxg template.
- An authorized run tests it against a target you own.
- The result is recorded in guardlink’s hypothesis ledger.
- It is rechecked when the code under it changes.
Reading code never confirms a claim. Only a reproduced run does. See the verification loop.
Free and Team plans
Section titled “Free and Team plans”| Plan | For | What it includes |
|---|---|---|
| Free (researcher) | one person, in any number of repositories, each on its own | the research loop: context, annotate, templatize, env, prove, recheck, report and watch. No licence, no account |
| Team (per developer; Solo is one seat) | individuals and teams | adds the pre-PR gate and autopilot, the CI and nightly workflows, scheduled upkeep, multi-repository workspaces, campaigns, SAST and the code graph |
Signing in is optional. bravos account activate starts a 30-day Team trial
with a free account and no card. See
how licensing works.
The dashboard combines the threat model and its design rules with a Team Code graph tab in one self-contained page.
Start here
Section titled “Start here”These pages link into cxg and guardlink
Section titled “These pages link into cxg and guardlink”What an annotation means, what a cxg flag does, or what a SARIF field holds is
already documented under /guardlink/ and /cxg/,
generated from those products’ own releases. A page here covers what Bravos
adds and links out for the rest.
Availability, as of 17 November 2026
Section titled “Availability, as of 17 November 2026”These pages describe Bravos 1.18.0, installed from npm.
| Component | Version | Where it is |
|---|---|---|
bravos CLI |
1.18.0 | public, on npm: npm install -g @bugb-technologies/bravos |
guardlink |
2.7.1 | public, on npm; bravos setup installs it |
cxg |
1.9.2 | public crate; bravos setup builds it with cargo |
The code graph (codegraph-build, codegraph-mcp) |
0.1.1, graph format 15 | Team, with a 30-day no-card trial. bravos setup downloads it through your licence and checks its signature |
The SAST analyzer (bravos-cli) |
0.6.2 | Team, with a 30-day no-card trial. bravos setup downloads it the same way |
| The plugins | 1.1.0 | public, Bugb-Technologies/bravos-plugins: Claude Code, Codex, Copilot, VS Code, Gemini CLI, Antigravity, Cursor. 14 skills; see Choose a Bravos installation |
| The PR gate action | 1.0.0 | public, GitHub Marketplace: uses: Bugb-Technologies/bravos-action@v1 |
Source
Section titled “Source”- CLI
- @bugb-technologies/bravos on npm; source is private
- Threat model
- Bugb-Technologies/guardlink
- Pentests
- Bugb-Technologies/cert-x-gen
- Licences and trials
- account.bugb.io

