Company profile · 4 min read
Agentic Fabriq: Identity and policy at every agent tool call
Agentic Fabriq gives AI agents identity, delegated permissions, credential handling and audit trails across business tools and customer accounts.
Cofounder of Lazyweb (a16z Speedrun). Previously grew subscriptions and built native ads at Duolingo. Writes First1000 (90,000+ subscribers).
Published · Updated
Agentic Fabriq is an identity, permissions and audit layer for AI agents. It fits a company moving from demos to agents that act for employees or customers across SaaS tools, databases and MCP servers.
What it does
Fabriq’s current enterprise site gives each agent an identity, carries the acting user through each request, enforces resource-level policy and records the result in a shared audit trail. The agent routes through Fabriq instead of holding provider tokens in its own context. Integrations listed publicly include GitHub, Gmail, Google Drive, Microsoft Teams, Notion and Slack.
The product has two distinct buyer paths. For an internal agent, the company needs per-user permissions, approvals, SSO and organization-wide audit. For a customer-facing product, each end user connects an account and Fabriq brokers the credential without the product holding the token. The developer page describes a hosted MCP gateway with 260-plus tools, OAuth per end user and an audit record for each tool call.
Pricing is unusually concrete. The pricing page shows Free at $0 for 20,000 calls per month, Launch at $50 for 200,000 calls and Growth at $300 for 2M calls. Additional calls on paid plans are listed at $2 per 10,000. Enterprise is custom for employee-wide governance, approvals, SSO, risk review and an organization audit trail. Every request through Fabriq—including authentication, tool access and credential retrieval—counts as a call, so the workload model matters.
Founder context and tradeoffs
The YC profile identifies Paulina Xu and Matthew Xu as founders. It describes Paulina’s AI and physics work at MIT and Matthew’s AI and math background. The launch explains the “Okta for agents” thesis: human identity systems do not answer who an autonomous agent acted for, what it could access or why a request was allowed.
The tradeoff is centralization. The team gains a single policy and audit layer, but must decide whether managed cloud storage fits its credential and data boundary. Fabriq says a custom plan can run in the customer’s AWS or GCP cluster; that option and the exact integrations need diligence.
Editorial take
I would shortlist Fabriq for an agent product with real delegated permissions and more than one integration. The free and Launch tiers make a narrow developer test accessible. I would not add it to a prototype that never touches user data; the value appears when identity, revocation and audit become part of the product promise.
Quick facts
| Field | Sourced detail |
|---|---|
| Product | Agent identity, permissions, credential handling and audit |
| Buyers | Internal-agent teams and customer-facing agent products |
| Public pricing | Free 20K calls; Launch $50/200K; Growth $300/2M; Enterprise custom |
| Developer surface | Hosted MCP gateway, OAuth brokering and 260-plus tools claimed |
| Main gate | Identity model, call economics, data residency and integration coverage |
Sources checked
| Source | Checked |
|---|---|
| YC profile | 2026-09-19 |
| Fabriq enterprise | 2026-09-19 |
| Fabriq pricing | 2026-09-19 |
| Fabriq developer platform | 2026-09-19 |
| Agentic Fabriq launch | 2026-09-19 |
Cohort context
Agentic Fabriq is listed in Winter 2026. In our 2026-09-18 directory snapshot, 126 of 199 listed companies in that cohort have YC’s primary industry label B2B (63.3%). This is a current-directory comparison, not an original intake count or a performance ranking. Nine-cohort dataset.
Public website snapshot
Observed 2026-09-19T16:20:00.523Z in raw homepage HTML. This records visible metadata and advertised links, not agent execution or product quality.
| Signal | Homepage observation |
|---|---|
| Product description metadata | Observed |
| Canonical link | Observed |
| H1 or H2 heading | Observed |
| Typed structured data | Observed |
| Docs/developer link | Observed |
| Pricing link | Observed |
| llms.txt link | Not observed in this response |
| Markdown alternate | Not observed in this response |
Public observations · Collection method. Missing links here do not establish that a capability or file is absent elsewhere.
About the author
Cofounder of Lazyweb (a16z Speedrun). Previously grew subscriptions and built native ads at Duolingo. Writes First1000 (90,000+ subscribers).
I cofound Lazyweb and publish Mudpie. This is an owner-written publication, not an independent testing organization. Research notes distinguish observations, sourced reporting and editorial judgment.
First1000 ↗ · X ↗