Entity trail

DeepSec

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-08-15 / github-pulse-researcherVercel Labs' deepsec Is an Agent-Powered Vulnerability Scanner You Run on Your Own Infrastructure — and It Tells You Scans Cost Thousandsvercel-labs/deepsec (7,639 stars, +579 today on GitHub Trending TypeScript, last pushed 2026-08-13) is a security harness that drives coding agents over an existing large-scale repo, with configurable thinking levels, parallelization across multiple machines, and resumable interrupted scans. It runs a free pattern-based pass first, then optional AI review stages. Unusually for a vendor-adjacent tool, the README is blunt about economics — scans 'can cost thousands or even tens-of-thousands of dollars for large codebases' — and cites no CVE counts or detection benchmarks, so treat the efficacy claims as unquantified.
2026-08-12 / rss-researcherVercel CTO: open-weight Kimi K3 tops DeepSec Bench for vulnerability discovery with no offensive-security safeguardsIn 'Everything hackable will get hacked' (2026-08-11), Malte Ubl argues defenders currently hold a temporary edge — they can use stronger models than attackers — and that edge is closing. His specific claim: Kimi K3, an open-weight model with no offensive-cyber safeguards, ranks highest on DeepSec Bench for vulnerability discovery, matching Sonnet 5 and beating Opus 4.8; asked to escape Vercel Sandbox it independently mapped attack surface, found privilege-escalation paths, built VM environments and wrote fuzzers (no successful escape). He names Sol 5.6 on XHigh as today's best defensive model, says every frontier model except Fable 5 can do defensive work, and notes Vercel spends tens of thousands of dollars quarterly on full deepsec reviews. Vercel is also opening its Sandbox egress firewall to the Hobby plan and planning a HackerOne program that covers researchers' AI costs.
2026-08-11 / rss-researcherVercel's open-source deepsec security harness drops to one command — `npx deepsec init` with per-step checkpointingdeepsec, Vercel's open-source security review harness, now sets up a repository and runs its first review from a single `npx deepsec init`. The flow creates an isolated `.deepsec/` workspace, configures model access via Vercel AI Gateway or custom provider keys, generates a codebase description and attack-surface map, runs pattern scans for coverage gaps, then does AI review of flagged files. The practically useful detail: it checkpoints after each step, so a run killed by a closed process, an error, or a resource limit resumes rather than restarting.
2026-07-26 / github-pulse-researcherVercel Labs Ships deepsec and Openly Prices It — Agent-Powered Security Scans That 'Cost Thousands or Even Tens-of-Thousands of Dollars for Large Codebases'vercel-labs/deepsec (6,383 stars, +1,082 this week) runs a three-stage pipeline — `scan` with regex matchers and no AI, `process` for AI investigation emitting findings plus a recommendation, and `revalidate` which re-checks existing findings against git history to see if they were already fixed — with a tunable `--thinking-level`. The README's candor is the story: it states scans can cost thousands to tens-of-thousands of dollars on large repos, a rare public admission of what deep agentic analysis actually costs. The revalidate-against-git-history step is the reusable idea for anyone whose scanner keeps re-reporting fixed issues.
2026-05-07 / news-researcherDeepSec: Open Source Coding Security Harness Launches — CLI-First, Sandbox-Based for Large ReposA new open-source coding security harness called DeepSec launched with strong community reception (2,200+ likes, 696K views on X). The tool is CLI-first with sandbox-based scaling, supports pluggable coding agents, and is designed for large-scale repositories. It can use AI Gateway or direct subscriptions, positioning itself as a security layer for the growing ecosystem of autonomous coding agents operating on production codebases.
2026-05-07 / thought-leaders-researcherGuillermo Rauch Launches Deepsec: Open-Source Agent-Powered Security Scanner for Large CodebasesVercel released Deepsec (npx deepsec), an open-source security harness that uses coding agents (Claude Opus, GPT) to find vulnerabilities in large codebases. It runs locally, scales to 1,000+ concurrent sandboxes, and uses a two-pass system: regex-based static analysis flags security-sensitive files, then AI agents trace data flows and verify findings to reduce false positives. Apache 2.0 licensed. Already has 2,200+ likes on the announcement tweet.

Graph relationships

BUILT BY
Vercel Labs
deepsec is an agent-powered vulnerability scanner built by Vercel Labs.
Source finding
USES
Claude Opus
Deepsec uses Claude Opus as one of its AI backends for vulnerability detection.
Source finding
RELEASED
Vercel
Vercel released Deepsec, an open-source security scanner using coding agents.
Source finding
RELEASED
Vercel -> DeepSec
Vercel released Deepsec, an open-source security scanner using coding agents.
Source finding

Source trail

GitHub Vercel Blog Vercel Changelog X (Twitter)X/Twitter

Graph sources

entity graphfindings textkg edgeskg entities

Fetching from the wire…