Enact is a critical piece of the AI agent stack, specifically addressing the 'Execution' and 'Safety' layers. As agents move toward economic agency—the ability to buy, sell, and trade on behalf of users—they require a trustless way to manage capital. Enact provides the onchain 'constitution' that limits what an agent can do, preventing a rogue or hallucinating model from draining a wallet.
For developers building in the agent ecosystem, Enact is a foundational tool for moving from demo to production. It allows for the creation of agents that are autonomous but bounded, using session keys and multisig protocols that are specifically designed for machine interactions rather than just human-to-human coordination. By integrating with major frameworks and providing an easy-to-use SDK, Enact is pushing the 'Agentic Finance' sector forward as a safer, more auditable field.
As AI agents transition from simple text generation to executing complex financial operations, the industry faces a fundamental trust gap. Giving an LLM direct access to a private key is a recipe for catastrophic loss, yet requiring a human to sign every micro-transaction defeats the purpose of autonomy. Enact is a infrastructure project building a middle layer that solves this by moving the 'permission' logic from centralized servers to the blockchain itself. They provide a non-custodial framework where teams can define granular, programmable rules that govern how agents interact with onchain capital.
Enact is currently in beta and is launching its mainnet presence on Tempo. The platform allows users to set spending limits, define specific approval thresholds, and automate rules directly within smart contracts. Because these policies live on the blockchain rather than on Enact’s own servers, there is no single point of failure. This architecture ensures that even if an agent is compromised or experiences a logic error, it remains constrained by the hard-coded parameters of the onchain policy.
The technical implementation of Enact centers on the enact-wallet SDK and a dedicated CLI. This setup allows developers to integrate agentic wallets into existing AI frameworks with minimal friction. A key feature of the system is support for session-scoped permissions. Rather than granting an agent permanent access to a vault, developers can issue temporary permissions for specific tasks or time windows. This is combined with flexible multisig controls, where both human signers and other AI agents can be added as required participants for certain transaction types.
Security is further bolstered by built-in key rotation and a commitment to non-custodial principles. Enact never holds user funds; it merely provides the enforcement logic that sits between the agent and the capital. For teams managing shared assets, this creates a transparent audit trail. Every action taken by an agent is recorded on the blockchain, providing a verifiable history that is critical for compliance and debugging in automated financial systems.
Enact is not building in a vacuum. The platform is designed to connect with the tools and models that developers are already using. Existing integrations include Anthropic’s Claude and OpenClaw, as well as a clean CLI interface for terminal-based agents. This strategy places Enact at the 'Action' or 'Execution' layer of the AI agent stack. While the LLM handles the reasoning and decision-making, Enact handles the authorization and safety checks.
By focusing on 'agentic finance,' Enact is betting that the future of decentralized finance (DeFi) will be dominated by autonomous signers. Their infrastructure addresses the practical realities of this shift—specifically the need for scaled payments, automated trading, and risk controls that run 24/7 without human intervention. As the ecosystem matures, the presence of a dedicated authorization layer like Enact may become the standard requirement for any production-grade agentic financial application.
Programmable non-custodial infrastructure for agentic onchain capital management.
Enact is hiring.