Ledgit

LEDGIT is a CLI-first compliance toolkit that brings enterprise-grade trust and auditability to fleets of AI agents. The core problem it solves is one of the biggest blockers to real-world agent adoption: “Your agent just executed a high-value action or moved real money — can you prove a human actually reviewed and authorized it?” LEDGIT enforces a secure human-in-the-loop process:

An agent proposes an action (payments, token swaps, role changes, contract upgrades, config updates, etc.). A human reviews clear details and approves directly on a Ledger hardware device Upon approval, the action (e.g. real token transfer) is executed on Hedera Testnet or mainnet. The full event — proposal, Ledger signature, payment receipt, and metadata — is immutably recorded in chronological order on a dedicated Hedera HCS topic Each agent (or business workspace) gets a clean ENS identity (e.g. trader-a.acme.ledgit.eth). ENS text records act as the discovery layer, pointing to the HCS topic and latest sequence

Logged actions are configurable via config.json, allowing enterprises to define their own high-stakes operations. A ledgit dashboard command opens a clean visual timeline of the entire audit trail. The result is a tamper-proof, cryptographically verifiable chain of custody that compliance teams, auditors, and regulators can trust instantly. LEDGIT turns powerful but opaque AI agents into accountable, production-ready systems for fintechs, DAOs, and regulated businesses.

Live site and try here: https://ledgitdash.vercel.app

LEDGIT is built as a Node.js / TypeScript CLI using Commander.js, making it lightweight and easy for both humans and agents (Claude, OpenClaw, etc.) to invoke via shell or tool calling. Key Integrations:

Ledger: @ledgerhq/hw-transport-node-speculos-http + Speculos emulator for realistic hardware signing or connect an actual Ledger device. Humans see human-readable action details on the device screen before approval.

Hedera: Hedera JS SDK + Agent Kit for executing real token transfers/payments on testnet. Every approved action is submitted as a structured (often encrypted) message to a per-agent HCS topic for immutable, ordered audit logs.

ENS: viem + ENSjs for subname management and text record updates (storing HCS topic ID and latest sequence number). Bring your own ENS for registering agent trails. Instead of some random account id, you can now quick lookup your data from ENS names that make it verify easy to see whose agent did what. XXXXXXX becomes chris-agent.acme-co.eth giving each agent it's own readable, and potentially human-tied, identity.

The flow is: ledgit propose → Ledger approval → Hedera payment execution → HCS record → ENS update. We added a configurable config.json system so new action types can be added without touching core code. For visualization, the ledgit dashboard command dynamically generates and opens a clean Tailwind HTML timeline.

Notable hack: We built a forced “unauthorized example” transaction in the demo as contrast to show the risk of running agents without LEDGIT.

Everything is open source with clear setup instructions, --json output for agent integration, and a 3-minute demo video showing a full vendor payment from proposal to verifiable audit trail.

Created as a net new build: https://github.com/cbonoz/ledgit