Stytch

Developer-first authentication/CIAM platform (passwordless, MFA, device-fingerprinting fraud) that has pivoted hard into AI-agent identity — turning your app into an OAuth provider for agents and MCP servers. Acquired by Twilio in Nov 2025.

One-liner — An API-first authentication and fraud platform for developers that now also does AI-agent identity: letting your app issue scoped, revocable OAuth tokens to agents and MCP servers, and detecting agentic traffic.

Categories — tool-identity-integration, non-human-identity

What it does

Stytch started as a developer-focused CIAM (customer identity and access management) platform: API-first, passwordless auth (email/SMS magic links, one-time passcodes, social sign-in, WebAuthn/passkeys, biometrics), session management, MFA, and a fraud/abuse layer built on device fingerprinting. The pitch was that auth should be embeddable building blocks for developers rather than a hosted “widget.”

Since 2025 its center of gravity has shifted to identity for AI agents. Two pieces matter for this wiki:

• Connected Apps — turns the customer’s own application into an OAuth 2.0 / OIDC authorization server, so AI agents and MCP servers can be granted scoped, auditable, revocable access to user data. It handles the full token lifecycle (issue/validate/refresh/revoke), consent screens, and human-in-the-loop / IT-admin approval, with the principle that an agent “inherits only the permissions the user already has.” This is the non-human-identity angle: delegated, least-privilege access for agents acting on a user’s behalf.
• IsAgent / agent fraud detection — IsAgent (launched 2025-08-01) is a lightweight front-end component that flags AI-agent and programmatic traffic (via User-Agent + TLS fingerprints, plus optional cryptographic self-identification) so sites can route agents to a tailored experience. Stytch is explicit that IsAgent is a “client hint, not a security guarantee”; for actual abuse prevention it points to its backend Device Fingerprinting product (intelligent rate limiting against scraping, prompt-injection-driven abuse, and compute abuse).

Where it sits in the stack

Primary category: tool-identity-integration (the agent→app/SaaS authentication-and-connector slot in the model/prompt layer); secondary: non-human-identity. Stytch is infrastructure a developer embeds, not a control a CTO buys off-the-shelf to govern their own AI usage — that distinction matters for the hedge-fund lens below.

Lethal-trifecta role: it mainly helps on the egress leg (scoped, revocable tokens bound what an agent can read/write through your app, instead of handing it a user’s full session) and partly on untrusted-input (IsAgent/fingerprinting + rate limiting filter malicious automated traffic, including prompt-injection-driven abuse). It does not classify or stop sensitive-data content itself. In trust-zone terms it governs the authorization boundary between an agent and a first-party application.

Deployment & architecture

• API + SDKs — REST/JS/backend SDKs plus pre-built UI components; “headless” option for full control. No inline proxy; it sits in the app’s auth/authorization path.
• MCP-native — native support for the Model Context Protocol with dynamic client registration, so remote MCP servers (connecting Claude, ChatGPT, etc.) get OAuth out of the box; deployment guides for Cloudflare Workers and Vercel edge.
• OAuth/OIDC authorization server — Connected Apps; Authorization Code Flow with PKCE for first-party apps, third-party integrations, and CLI tools.
• Fraud layer — Device Fingerprinting backend + IsAgent front-end component; intelligent rate limiting.
• Key integrations: standard IdPs/social providers, MCP clients, and (post-acquisition) Twilio’s communications/identity reputation data.

Positioning & differentiators

Stytch is known as a developer-first, API-first auth alternative to hosted CIAM, and in 2025 it moved early and aggressively on agent/MCP authentication, branding itself an “identity platform for AI agents.” Versus nearest neighbors:

• descope — the closest comparison: also a developer-centric CIAM that has leaned hard into agentic/non-human identity and MCP auth. Stytch differentiates on its device-fingerprinting fraud heritage and (now) Twilio’s reputation data.
• workos — more focused on enterprise-readiness features (SSO/SCIM/directory sync) that B2B SaaS bolt on to sell upmarket; overlapping but more “enterprise plumbing,” less fraud/consumer-auth and (historically) less agent-fraud tooling.
• okta (Auth0) — the incumbent identity/CIAM platform; broader and more enterprise, with its own agent-identity initiatives. Stytch positions as lighter-weight and more developer-native.
• cyberark — comes at non-human/agent identity from the enterprise secrets/privileged-access side, governing workloads centrally, rather than embedding OAuth into an app a developer ships.

Ownership, funding & M&A

Acquired by Twilio. Definitive agreement announced 2025-10-30 (alongside Twilio’s Q3 2025 results); deal completed 2025-11-14. Terms were not disclosed. Twilio’s stated rationale is augmenting its platform with modern auth and AI-agent authentication/verification, pairing Stytch’s identity stack with Twilio’s communications/reputation data. Confirmed against Twilio’s own announcement and Stytch’s changelog → ownership: acquired, confidence high.

Pre-acquisition funding: $30M Series A (2021, led by Thrive Capital) and a $90M Series B (2021-11, led by Coatue) at a $1B valuation = $120M in disclosed rounds; aggregators (Crunchbase/Tracxn) list cumulative funding around $126M–$146M, which is not vendor-confirmed. Founded 2020 in San Francisco by Reed McGinley-Stempel (CEO, ex-Plaid) and Julianna Lamb (CTO); some sources also list Mark Cunningham as a co-founder.

CTO / hedge-fund lens

Day-2 at most, and for most funds: not a buy at all. Stytch is developer infrastructure you embed when you are building an application (or an MCP server / agent-facing API) and need to authenticate users and grant agents scoped access. A hedge fund that is a consumer of AI tools — running ChatGPT/Claude Enterprise, governing staff usage, doing entitlement-aware RAG — does not buy Stytch; that’s why hedge_fund_fit: low.

It becomes relevant only if the fund builds its own customer- or agent-facing software (e.g. an investor portal, an internal agentic app exposed via MCP) and wants delegated, revocable, least-privilege agent access rather than handing agents broad credentials. In that narrow case its Connected Apps / MCP-auth model is a sensible control on the agent-authorization boundary. No direct SR 11-7 / model-risk relevance. Also weigh the Twilio acquisition: roadmap and pricing now sit inside a larger platform — a diligence point, not necessarily a negative.

Competitors / alternatives

descope, workos, okta, cyberark, composio, arcade, stackone

Open questions / to verify

• Vendor-confirmed cumulative funding (aggregators disagree: $126M vs $146M; disclosed rounds total $120M).
• Confirm Mark Cunningham as co-founder from a primary source (press names McGinley-Stempel and Lamb consistently; Cunningham appears in aggregator profiles).
• Post-acquisition product status: will Connected Apps / IsAgent persist standalone or fold into Twilio’s identity offering, and any pricing/availability changes.
• Independent (non-marketing) validation of device-fingerprinting / agent-detection accuracy.

Sources

• Twilio Acquires Stytch, an Identity Platform for AI Agents — fetched 2026-06-28 — supports: acquisition by Twilio, announced 2025-10-30 / completed 2025-11-14, terms undisclosed, agent-identity rationale; confidence: high (acquirer primary).
• Stytch raises $90M Series B at $1B valuation (TechCrunch) — fetched 2026-06-28 — supports: $90M Series B 2021-11 (Coatue), $30M Series A, $1B valuation, SF HQ, API-first passwordless product, developer target; confidence: high.
• Stytch Agent Ready / AI agents & Connected Apps docs — fetched 2026-06-28 — supports: Connected Apps OAuth/OIDC for agents, MCP-native auth, scoped/revocable tokens, consent, agent fraud/rate-limiting, SDK+API deployment; confidence: med (vendor).
• Introducing IsAgent (Stytch blog) — fetched 2026-06-28 — supports: IsAgent (2025-08-01) agent detection via UA+TLS fingerprints, “client hint not a security guarantee,” Device Fingerprinting for fraud; confidence: med (vendor).
• Founders/founding year (2020; McGinley-Stempel, Lamb, Cunningham) and ~$126M–$146M cumulative funding: Crunchbase/Tracxn aggregators (not cached) — confidence: low/med.

History

• [2026-06-28] Stub created from seed registry.
• [2026-06-28] Researched; established founded 2020 (San Francisco), founders Reed McGinley-Stempel (CEO) & Julianna Lamb (CTO) (+ Mark Cunningham per aggregators), developer-first CIAM with $90M Series B (2021, Coatue, $1B valuation). Corrected ownership from independent to acquired: Twilio acquired Stytch, announced 2025-10-30, completed 2025-11-14 (terms undisclosed), confirmed via Twilio’s primary announcement — confidence high. Documented the agent-identity angle (Connected Apps OAuth-for-agents, MCP-native auth, IsAgent/device-fingerprinting). Positioned vs descope/workos/okta/cyberark; set hedge_fund_fit low (developer infra, not a CTO buy). status: researched, confidence medium. 4 sources cached.