deepsec README

The project's README from github.com/vercel-labs/deepsec. As one markdown file: README.md.

deepsec

deepsec is an agent-powered vulnerability scanner that you can run in your own infrastructure, optimized to perform on-demand review of all code in existing large-scale repos.

deepsec is designed to surface hard-to-find issues that have been lurking in applications for a long time. It is configured to use the best models at maximum thinking levels (tunable via --thinking-level, see models), meaning scans can cost thousands or even tens-of-thousands of dollars for large codebases. Our customers have found the cost worth it for how quickly they were able to patch vulnerabilities that would have otherwise gone unfixed.

For large codebases, work fans out across worker machines in parallel. If a run is interrupted or errors out partway through, just re-run the same command — deepsec picks up where it left off, skipping files it already analyzed and only investigating the rest.

Get started

From the root of the repository you want to scan:

npx deepsec init

The command guides you through everything. It asks you to pick an AI model (with benchmark scores and prices to compare) and how to pay for model usage — your own OpenAI/Anthropic API key, or Vercel AI Gateway — and then works unattended: it studies your codebase, scans it, and runs the AI review. The only thing it adds to your repository is a .deepsec/ folder where all of its state and findings live.

If the run is interrupted for any reason — Ctrl-C, lost connection, a spending limit — run npx deepsec init again and it continues where it left off. To cap what a run may spend or how long it may take:

npx deepsec init --max-cost-usd 100 --max-duration 2h

When the scan finishes, get a readable report:

cd .deepsec
pnpm deepsec export --format md-dir --out ./findings

For later scans, work from inside .deepsec/:

pnpm deepsec scan        # fast pattern scan, free
pnpm deepsec process     # AI review of new candidates
pnpm deepsec revalidate  # optional, cuts false-positive rate
pnpm deepsec export --format md-dir --out ./findings

The getting started guide covers all of this in more detail, including using your own OpenAI or Anthropic API key and running from CI or a coding agent.

Docs

After initialization, agents can read the exact documentation matching the installed CLI at .deepsec/node_modules/deepsec/SKILL.md and .deepsec/node_modules/deepsec/dist/docs/. Setup errors expose these as absolute machine-readable paths.

AI provider

By default, deepsec routes model calls through Vercel AI Gateway, which gives access to every major model without provider-specific keys. You can instead bring your own key — OpenAI, Anthropic, or a custom HTTPS provider — by passing --model-auth direct with --ai-provider and --ai-api-key-env to init; no Vercel account is needed in that mode. Deepsec only ever stores the name of the environment variable holding your key, never the key itself. See project link and credentials for the full reference.

If a process or revalidate run halts because the upstream credential ran out of quota or credits, deepsec stops gracefully and tells you where to top up. Re-run the same command afterward and it picks up where it left off.

Distributed execution (optional)

Large monorepos can fan work across Vercel Sandbox microVMs:

pnpm deepsec sandbox process --project-id my-app --sandboxes 10 --concurrency 4

Setup already verified the Vercel connection, so this needs no extra onboarding. The local working tree is tarballed and uploaded; .git is excluded. Model credentials remain host-side and are injected only at the selected egress host.

Security model of deepsec itself

Treat deepsec like a coding agent with full shell access on the enviroment that it is running on. It is designed to run on trusted inputs (your source code) but you may still be concerned about prompt injection due to external dependencies or vendored code.

Running on a sandbox (see above) does limit the potential exposure substantially:

Workflow reference

Command What it does
scan Find candidate sites with regex matchers (fast, no AI)
process AI investigation; emits findings + recommendation
process --diff PR-mode: scan + investigate only files changed in a diff
triage Lightweight P0/P1/P2 classification (cheaper model)
revalidate Re-check existing findings; checks git history for fixes
enrich Add git committer info + (with a plugin) ownership data
report Markdown + JSON summary for one project
export Per-finding JSON or directory of markdown files
metrics Cross-project counts: severities, vulns by type, TPs
status Snapshot of the project mirror
sandbox <cmd> Run any of the above on Vercel Sandbox microVMs

License

Apache 2.0. See LICENSE and NOTICE.

deepsec (vercel-labs/deepsec)

Searching Database…

8.2K ·GitHub
Deepsec
What is this?Deepsec is an agent-powered vulnerability scanner that you run in your own infrastructure to do on-demand security review of an entire large codebase, surfacing hard-to-find issues that have been lurking for a long time.
Is this a library I import or a CLI tool I run?It is a command-line tool invoked with npx, not a library you import into code; it writes all its state and findings into a .deepsec/ folder in your repo.
How do I run my first scan?From the root of the repo you want to scan, run 'npx deepsec init'. It walks you through picking an AI model and a payment method, then scans unattended and keeps all its state in a .deepsec/ folder.
Do I need my own OpenAI or Anthropic API key?Not necessarily — by default deepsec routes model calls through Vercel AI Gateway, giving access to every major model without provider-specific keys. You can instead bring your own OpenAI, Anthropic, or custom HTTPS key via --model-auth direct.
How expensive can a scan get?Deepsec is configured to use the best models at maximum thinking levels, so a full scan of a large codebase can cost thousands or even tens of thousands of dollars in model usage.
Who is this built for?It's built for engineering/security teams who need to audit large, existing codebases for long-standing vulnerabilities and are willing to pay for heavy AI model usage to find them.
Can malicious code it scans exfiltrate my API keys via prompt injection?Deepsec runs with full shell access like a coding agent, so vendored or external code could attempt prompt injection, but running it in a sandbox limits exposure: API keys for coding agents are injected outside the sandbox and worker sandboxes can only reach coding-agent hosts over the network.
What happens if a scan gets interrupted?If a run stops due to Ctrl-C, a lost connection, or a spending limit, re-running the same command resumes it, skipping files already analyzed and only investigating the rest.
How does deepsec handle very large monorepos?Work fans out across worker machines, optionally on Vercel Sandbox microVMs, where your local working tree (minus .git) is tarballed and uploaded so analysis runs in parallel.
What commands make up the deepsec workflow?Beyond scan (fast regex matching) and process (AI investigation), there's triage for cheap severity classification, revalidate to recheck old findings against git history, enrich for ownership data, and report/export/metrics to summarize results.
What are similar projects?strix, Airsec, 0, qwen-code, open-swe, codex-security, mantis, open-code-review
Which topics is it in?vulnerability-scanner, code-review, static-analysis, sandbox, ci, security-tools, anthropic, cloud, coding-agent, ai-agent, openai, agent
Creator
vercel-labs
Size
1.7 MB
Stars
8.2K
Updated
CommitSep 29
Age
5mo
Commits · 90 Days