Startups building payment and identity rails for AI agents transacting autonomously
August 26 at 09:21 · $0.111 total
Agentic Commerce Rails — Sourcing List
1. Skyfire — Payment network letting AI agents hold funds and pay for APIs, data, and services autonomously with spend controls. Fits as near-pure-play on the thesis; seed-stage, backed by a16z crypto and Coinbase Ventures, early KYA (know-your-agent) identity layer bundled with payments.
2. Payman AI — Platform enabling AI agents to pay humans (and vice versa) with programmable guardrails and human-approval workflows. Early-stage (backed by Visa and Coinbase Ventures); interesting inversion — agents as payers of human labor, a mostly unclaimed niche.
3. Nekuda — SDK giving agents secure access to user payment credentials so they can complete checkout on the user's behalf. Very early (seed, Madrona-backed); attacks the "agent hits a checkout wall" problem directly, complementary to card-network agentic tokens.
4. Catena Labs — Building a regulated, AI-native financial institution designed for agent-driven transactions. Founded by Circle co-founder Sean Neville; raised ~$18M (a16z crypto). Strong founder-market fit signal; bet is that agents need agent-native banks, not retrofitted rails.
5. Crossmint — Wallet and stablecoin infrastructure APIs, increasingly positioned around wallets for AI agents. Beyond seed but still underexposed; agent-wallet SDK traction with agent frameworks makes it a picks-and-shovels play on agent-held funds.
6. Paid (paid.ai) — Billing, pricing, and margin infrastructure for AI agent companies (outcome-based pricing, usage metering). Founded by Manny Medina (Outreach); adjacent angle — the monetization rail for agents rather than the transaction rail; early but strong founder signal.
7. Kite — Blockchain/L1 purpose-built for agent identity and stablecoin micropayments between agents. Raised from PayPal Ventures and General Catalyst; represents the crypto-native thesis that agent-to-agent microtransactions need new settlement rails.
8. Anon — Delegated authentication infrastructure letting agents securely act within users' accounts on sites without APIs. Seed-stage; identity/auth delegation is the unglamorous prerequisite for agentic commerce, and Anon is one of the few dedicated players.
9. Natoma — Non-human identity management platform, extending workforce IAM patterns to AI agents and service accounts. Early (seed); rides the enterprise wave of "who is this agent and what is it allowed to do" — governance layer the thesis requires.
10. Aembit — Workload and agent identity-and-access management (secretless auth between agents, apps, and APIs). Series A security company pivoting hard into agent IAM; slightly later stage but a credible identity-rail incumbent-in-the-making, still under most radars.
(Caveat: this space moves fast — stage/traction details should be re-verified at time of outreach.)
---
White-space gaps noticed
- Dispute resolution and liability rails for agent transactions. Everyone is building the "pay" leg; almost no one is building chargebacks, fraud adjudication, or liability attribution when an agent transacts incorrectly. Whoever becomes the "arbitration layer" for agentic commerce captures a Visa-like position.
- Reputation/credit scoring for agents. KYA verification exists at issuance (Skyfire, Kite), but there's no FICO-equivalent: a portable, cross-platform trust score based on an agent's transaction history that merchants and counterparty agents can price risk against. It's the missing primitive for agent-to-agent commerce at scale.
- Skyfire — Builds a payment network and “Know Your Agent” identity layer so autonomous AI agents can hold wallets, establish trust, and pay for services; fit: one of the clearest early pure plays, with a seed round and pilots aimed specifically at agent-to-business transactions.
- Nevermined — Provides programmable payment infrastructure, usage credits, and micropayments for AI agents and agent-to-agent services; fit: early startup explicitly positioning payments as the missing economic layer for autonomous AI, with a funded protocol and developer platform.
- Payman AI — Enables AI agents to pay people and service providers through controlled wallets and programmable payment workflows; fit: early, highly focused company built around the practical problem of agents compensating humans for completing tasks.
- Turnkey — Offers embedded wallet, key-management, and policy infrastructure that lets applications create and control wallets programmatically; fit: its threshold-signature and policy controls are well suited to agent wallets that need bounded autonomy rather than unrestricted private-key access; backed by a significant early-stage venture round.
- Crossmint — Provides APIs for embedded wallets, stablecoin payments, digital assets, and programmable accounts, including infrastructure designed for AI-agent use cases; fit: funded infrastructure company with production APIs and traction across developers, increasingly extending its wallet stack toward autonomous software.
- Lit Protocol — Supplies programmable key management and decentralized signing, allowing permissions, spending limits, and execution rules to be attached to wallets and identities; fit: an early crypto infrastructure company whose delegated-signing model is a strong primitive for agents acting without giving them full custody.
- Dynamic — Builds authentication and embedded-wallet infrastructure for applications, supporting multiple identity methods and wallet experiences; fit: its combination of user identity, wallet provisioning, and policy-controlled access can serve as the identity layer for agent and human-supervised transactions; venture-backed and already used by many developer-facing applications.
- Composio — Provides authenticated tool and API access for AI agents, handling OAuth, credentials, permissions, and integrations across business software; fit: early startup directly addressing agent identity and authorization—ensuring an agent can act as a scoped principal rather than using a developer’s unrestricted credentials.
- Nango — Offers a unified integration and authentication layer for software connecting to third-party APIs, including agent applications; fit: seed-stage infrastructure company positioned around OAuth management, credential refresh, and permissioned access—the “what may this agent do?” layer adjacent to payments.
- Pipedream — Provides “Connect” infrastructure for agent developers to authenticate users and let agents invoke third-party tools and APIs; fit: an established but still relatively young developer platform with broad integration coverage and a direct use case in giving agents delegated, auditable permissions.
White-space gaps noticed:
- Portable agent identity and reputation: There is no widely adopted equivalent of a passport or credit bureau for agents that binds an agent’s owner, permissions, operating history, liability, and reputation across payment networks and SaaS platforms.
- Autonomous payment risk and dispute infrastructure: The market lacks standardized controls for agent spending limits, merchant verification, fraud detection, refunds, escrow, and human override—especially for micropayments and agent-to-agent commerce across multiple currencies and chains.