CrowdStrike Veteran Loses Tracecat Role to Unknown Contributor.

What Gets You Past Tracecat's Screen

Listing Splunk SOAR on your résumé won't land an interview. The screen demands proof you've built agents that make security decisions inside a real toolchain, plus evidence you've done it in the open where anyone can inspect the code.

Chris Lo, the chief executive, built machine-learning tooling for security and fraud detection at PwC. Daryl Lim, the chief technology officer, shipped observability and AI infrastructure at AMD, Meta, and Maven Securities after computer-engineering degrees at Imperial College London. Their combined background means the screen weighs two things equally: you understand security operations (alert triage, threat hunting, incident response), and you can ship the infrastructure that lets an agent act on those operations without melting the production database.

The screen checks for: you have written Python that talks to a SIEM or EDR API and handled the pagination, rate limits, and schema drift that break naive integrations. You have designed a durable workflow on Temporal or an equivalent framework (retries, timeouts, saga compensation), and you can explain why a scatter-gather fan-out needs idempotent steps. You have implemented human-in-the-loop approval gates where a denied tool call stops the agent rather than triggering a retry. You have published an MCP server or consumed one, and you know the difference between a stdio-local server and a remote HTTP/OAuth endpoint. You have version-controlled prompts, skills, and workflow definitions in Git and treated them as code, not configuration. And you have done at least some of this in a public repository where a reviewer can read the diffs.

The founders' Y Combinator Winter 2024 batch gets mentioned in every intro deck, but the screen does not care about the badge. It cares whether you can operate at the velocity the repo demonstrates: a 1.0.1 patch two days after 1.0, then a 1.1.0 alpha series through late September shipping on October 9. That pace leaves no room for engineers who need a ticket to tell them what to refactor. The hiring window is two roles. The signal that gets you through is code you've already shipped in the same problem space, and the funnel mechanics shift in ways most applicants underestimate.

Why a Limited Slate Changes the Odds

Tracecat's engineering slate is two roles total: one Software Engineer (Full-stack) in New York, plus a Mid-Market Account Executive, and only the engineering role was added in the past seven days. The company's own salary band across two salaried roles is summarized below. That is the entire payroll footprint.

When a YC W24 startup with two founders opens a single engineering seat, the denominator is small enough that every resume gets human eyes, but the numerator is small enough that the bar clears only for candidates who can demonstrate the exact work the role demands: agentic automation, security operations fluency, and production-grade TypeScript/Python in a codebase that is already public.

A typical Series A security company might interview 50–100 engineers for five slots. Tracecat's ratio is closer to one slot for whatever inbound volume the YC brand and 1,958 LinkedIn followers generate. Pedigree (FAANG tenure, Stanford MS, prior YC founder status) becomes noise. Signal is a merged PR in the Tracecat repo, a written post-mortem of a SOAR playbook you built, or a live demo of an MCP server you wired to a SIEM.

The next engineer hired becomes employee three or four on the technical side. There is no onboarding cohort, no buddy system, no "ramp quarter." The hire ships on day one. The screen therefore optimizes for zero-ramp contributors: people who have already solved the class of problems Tracecat faces (human-in-the-loop approval flows, idempotent action execution, audit-log fidelity) and can prove it in a public fork. A candidate with three years at CrowdStrike but zero open-source commits loses to a candidate with one year at a no-name MSSP who contributed a connector to the Tracecat registry last month.

YC W24 amplifies inbound but does not widen the slot. The batch label draws applications from founders' networks, demo-day viewers, and the "YC job board" crowd, many of whom treat the application as a lottery ticket. Tracecat's screen treats it as a filter. The limited slate forces the team to reject qualified people who cannot show the work, not because they lack ability but because the cost of a false positive at this stage is existential. One bad hire at employee three rewrites the product trajectory. The odds are not "competitive" in the usual sense; they are binary. You either produce the artifact that maps to the open issue, or you do not clear the screen; that artifact has to live in a stack that looks nothing like generic DevOps.

AI-Native Security Is Not Generic DevOps

Tracecat runs on Kubernetes not because it likes containers but because its execution engine is Temporal, and Temporal needs a durable, horizontally scalable backend to persist every workflow and agent step at 22 million actions a day. That distinction — durable execution as a first-class requirement, not an afterthought — separates agentic SOAR from generic DevOps automation.

Traditional SOAR is playbook-centric. Teams wire fixed steps across integrations, then maintain those playbooks as tools and processes change. Tracecat keeps the SOAR platform open source and code-native. Security engineers build workflows, write Python, and manage cases. Agents handle the work that changes from alert to alert while policies gate sensitive actions. SOAR automates response with playbooks whose steps are defined in advance. Agentic SOAR adds agents that plan investigations, delegate specialized work to subagents, apply reusable skills, and call security tools through MCP servers. Those agents draw security context from case history and searchable table data, plus semantic search in connected tools that offer it. Deterministic workflows still run the repeatable paths. Audit logs record what agents do. Human-in-the-loop agent approvals gate sensitive actions.

MCP (Model Context Protocol) is the connective tissue. Tracecat hosts 50-plus MCP servers for security operations and exposes more than 100 tools through its MCP interface. Your coding assistant (Claude Code, Codex, Copilot) drives the entire platform through prompts. An agent that quarantines a host is a privileged actor; its prompts, tool calls, and decisions should stay in your Git, your VPC, and your audit logs, where your team can review them. The platform ships 100-plus pre-built connectors across SIEM, EDR, MDM, IdP, and email security via HTTP, SMTP, and gRPC. Integrations are Python packages or YAML templates in a repository you control, pinned to a commit with promote and rollback. New actions load without an image rebuild.

Human-in-the-loop is not a checkbox. Sensitive tool calls pause for explicit approval from a unified inbox, Slack, or email. Every approval is logged. The same agent preset runs in workflows, chat, and cases. Any tool can wait for a person's approval. This is not "human on the loop" observability; it is a gating mechanism built into the control flow, which itself supports loops, if-conditions, parallel subflows, and scripts in Python, Bash, and Ansible.

The stack reflects the domain. Backend: Python, FastAPI, Pydantic. Frontend: Next.js, TypeScript. Durable execution on Temporal. PostgreSQL for data storage. Sandboxed execution via nsjail, on by default in Kubernetes. GitOps for everything: workflows, agent presets, skills, table schemas, case configuration sync to a GitHub, GitLab, or Bitbucket repository your team owns. A push opens a pull request. Prompt versioning with diff, review, and rollback. Custom registry syncs custom Python scripts from your Git repo into Tracecat. Fine-grained access control: RBAC, ABAC, OAuth2.0 scopes for humans and agents. Multi-tenant workspaces isolate dev from prod. SAML and OIDC support with no SSO tax. Exportable audit logs feed your SIEM.

Deployment flexibility is a security requirement, not a convenience. Alerts carry hostnames, user identities, and raw evidence that many teams cannot send to a vendor tenant. Self-hosting the control plane — not just a runner — keeps that data inside your own network boundary. Open source self-hosts on Docker Compose or AWS Fargate. Enterprise adds a Kubernetes Helm chart and a managed cloud in the US or EU. The air-gapped deployment guide covers production Kubernetes, including self-hosted LLM inference. Applicants mapping experience should look for: production Temporal or comparable durable-execution systems; building or operating MCP servers or clients; writing Python integrations that ship as versioned packages; designing approval-gated control flows for privileged actions; running air-gapped or VPC-isolated control planes; and contributing to open-source security tooling where the ontology (OCSF, Tracecat's own capability taxonomy) shapes the integration surface. Generic CI/CD pipeline experience does not transfer. This is agentic security operations (triage, investigation, threat hunting, incident response) running on a platform your team owns, inspects, and extends. The product architecture recognizes only one credential: public contribution.

Open Source Contribution as the Primary Credential

Tracecat's entire product lives in public. The core platform ships under AGPL-3.0 with 6,043 commits logged as of October 18, 2026, and the repository shows 3.8k stars, 435 forks, and 57 contributors as of October 19. That contributor count matters more than the star total; n8n sits at roughly 160k stars but targets general automation, while Tracecat's 57 contributors reflect a specialized security-operations community building agents, workflows, and MCP integrations in the open. The project registered on SourceForge in November 2024 and reached version 1.2.1 by October 19, 2026, with commits landing weekly: a UI dialog fix on October 18, a workflow rerun feature on October 15, a PostgreSQL trigram index migration on October 13. The pace is visible, verifiable, and unforgiving of vaporware.

The hiring screen inherits that transparency. Tracecat's platform is built on GitOps: "Sync everything to Git" is not marketing copy but the actual architecture. Enterprise workspaces push those artifacts and case configuration to that repository the team owns. The open-source edition includes unlimited workflows, agents, cases, SSO, built-in roles, and organization audit logs. Anyone can clone the repo, spin up the stack on either platform, and start shipping integrations against the 100+ pre-built connectors and 50+ hosted MCP servers. The sandboxed execution layer (nsjail) and durable Temporal backend mean contributions land in a production-grade runtime, not a toy environment.

For an applicant, this changes what "experience" means. A résumé line about Splunk SOAR or Tines proves you operated a vendor's black box. A merged pull request adding a connector for a niche EDR, fixing a sandbox escape path, or extending the MCP server registry proves you understand the engine (FastAPI, SQLAlchemy, Pydantic, uv on the backend; Next.js, TypeScript, React Query, Shadcn UI on the frontend) and the security domain it serves. The 57 contributors include engineers from Saronic, NinjaOne, Depop, Lendi Group, DEVCOM, Wealthsimple, Acumen, CyberUp, and Neo, organizations running Tracecat in production, some in air-gapped federal environments. Their commits are reference implementations. Studying them is interview prep the company cannot gatekeep.

The Tracecat founding team, Daryl Lim and Chris Lo, built the product in public from day one. The Software Engineer (Full-stack) role at that band does not ask for a specific degree or a FAANG pedigree. It asks for evidence you can operate in this stack, in this domain, in the open. AIMultiple's 2026 open-source SOAR analysis explicitly cites GitHub activity (stars, contributors, commit velocity) as the practical starting point for evaluating a project's health. Tracecat's 57 contributors and weekly releases signal a living codebase where a competent engineer can ramp fast and contribute meaningfully before the first interview.

Community proof compounds. The platform's MCP layer lets coding agents (Claude Code, Codex, Copilot) turn prompts into workflows. A candidate who has built a custom MCP server for a security tool, published it, and documented the integration has demonstrated the exact loop Tracecat's customers run daily: prompt → automation → case resolution → Git sync. Testimonials on the company site describe one-person SOCs building IoC enrichment pipelines in a day using Tracecat MCP, the kind of leverage the startup sells. An applicant who can point to a similar public artifact, even a modest one, carries more signal than a certificate or a generic "AI automation" bullet point. The screen narrows further because the product itself is the interview. Breaking changes between releases are documented in the changelog; the team advises reviewing it before each update. An engineer who has navigated those upgrades in a self-hosted deployment — wrestling Temporal, PostgreSQL, and optional LLM infrastructure — has already done the job. The research notes this infrastructure burden as a constraint for smaller teams, which means the hiring bar selects for people who have already absorbed it. Public contribution is not a proxy for skill here. It is the only credential the product architecture recognizes, and the founders hire from a New York office with a two-role slate.

New York Base, Remote Reality, and the YC Label

Tracecat lists its headquarters as New York City across its Y Combinator profile, its Work at a Startup page, and the live job posting dated October 17, 2026. Crunchbase still shows San Francisco, a common lag for early-stage companies that incorporated in one city before planting the team in another. The October posting is explicit: "2 day on-site at our New York office" and "15 days of remote work flexibility per year." That is not a remote-first policy. It is a New York role with a narrow travel allowance. Candidates who cannot commit to two days a week in Manhattan should not apply.

The company was founded in 2024 and went through Y Combinator's Winter 2024 batch. That batch label (W24) appears on every directory listing. It signals a few concrete things: a $500,000 standard check (post-2022 terms), access to the YC network for hiring and later fundraising, and a demo-day deadline that forced a minimum viable product into the world. It does not signal product-market fit, security-domain credibility, or a differentiated technical moat. The founders built that credibility themselves. Lim's background runs through AMD, Meta, and a London quant trading firm before Tracecat; he holds bachelor's and master's degrees in computer engineering from Imperial College London. Lo's public footprint is thinner, but the pair have kept the same mission since demo day. "SOAR is not dead," Lo said in a March 2026 LinkedIn retrospective. "AI will increase the demand for build not buy," Lim said in the same post. "Most teams that have or want a SOAR do a lot more than just alert triage." They repeated those lines two years after batch.

Two investors show on Crunchbase: Pioneer Fund and Y Combinator. No Series A has been announced. The company describes itself as "a product-led open core company winning deals against competitors 100x their size" in the October software-engineer posting. That claim is backed by usage numbers the founders shared in September 2026: 50,000 downloads, 20 million automations per month in their cloud, and over 1,000 agentic workflows built to date. Growth, they said, is "spiking purely through word of mouth." Those are traction metrics, not vanity metrics; they map directly to the open-source product that engineers would be hired to extend.

The YC affiliation also shapes the hiring funnel in a specific way. YC companies at this stage typically recruit from the YC talent pool, alumni referrals, and the Work at a Startup board. That pool self-selects for founders and early employees who have already chosen the startup path. It does not self-select for security practitioners. Tracecat's screen — "full-stack generalist engineers who ship fast with an unrelenting bar for quality," "technical depth and software craftsmanship: we're an anti-slop AI-pilled company," "product intuition: you care about the 'why' as much as the 'how'" — is written for that audience. It assumes the candidate can write production Python, design APIs, and reason about distributed systems. It does not assume they know the difference between a SIEM and an EDR. That gap is exactly where the open-source contribution signal becomes the primary credential.

The equity ceiling is low for a seed-stage company, which suggests the founders are preserving ownership for a priced round or are calibrating to New York market rates rather than Bay Area norms. Founders Daryl Lim and Chris Lo are the only named principals. What the YC label does not do is lower the technical bar. The October posting's "slope over intercept" line — "we value learning velocity, grit, and unapologetically unique personalities" — is a culture cue, not a competence substitute. The screen still demands verifiable code in the agentic-automation domain. The New York office requirement means the talent pool is geographically constrained to the tri-state area or people willing to relocate. The 15-day remote allowance covers conferences, family events, or a quiet sprint week, not a distributed-work arrangement. For applicants outside that radius, the window is effectively closed unless they move. That constraint, combined with the two-role slate, makes the funnel narrower than the YC brand alone would suggest, and the report's scope ends at the screen.

What This Hiring Report Does Not Cover

No salary deep-dive

First-party board data shows a salary band across Tracecat's two salaried roles (detailed above). Zero G Talent's board data shows the Mid-Market Account Executive range tops out at $350K. Broad market surveys such as ISG Partners' 2026 engineering compensation report or peer-sourced datasets like BenchComp's anonymous submissions provide context this report intentionally omits.

No non-engineering roles beyond the two listed

As of the latest board ingest, Tracecat has exactly two open roles on Zero G Talent. The board does not surface openings in product management, design, developer relations, security research, or operations.

No competitive vendor comparison

Tracecat operates in a crowded category. CybersecTools lists 48 alternatives. FuturePicker publishes an in-depth comparison of Tracecat versus Tines, Shuffle, and TheHive. This report does not replicate those comparisons.

What remains in scope

The exclusion list above is not a disclaimer — it is a boundary. Inside that boundary, the report traces how a YC-backed, open-source-first security startup structures a two-role engineering screen around verifiable code rather than résumé keywords. The interview room isn't a conference room in Manhattan. It's the GitHub repo where a merged PR adding a Wiz connector or fixing a Temporal retry policy sits next to the founders' own commits — public, verifiable, and already doing the work the role demands.

Working in frontier tech? Zero G Talent tracks the openings: see every open Tracecat role, browse frontier tech jobs, the companies hiring, and the people building the field.