Trust gap
Final answers hide the decisions, retries, rejected approvals, and tool behavior that actually create risk.
Built for 2026 enterprise-agent hackathons
A payment-aware ledger for production-like agents: every tool call, approval gate, retry, error, x402 payment, and final outcome becomes inspectable run evidence.
Enterprise teams do not need another hidden transcript. They need durable operational evidence before agents can touch paid APIs, internal systems, or approval-gated workflows.
Final answers hide the decisions, retries, rejected approvals, and tool behavior that actually create risk.
Security and platform teams need searchable event streams, not screenshots of a chat window after the fact.
Agents calling paid APIs need cost attribution tied to the exact run, endpoint, provider, and outcome.
The build turns an agent workflow into a run ledger with dashboard inspection, exportable audit JSON, x402-style payment events, and Splunk HEC telemetry.
startRun, trackToolCall, trackApproval, trackPayment, and finishRun wrap existing agent code without locking it to a framework.
Express and SQLite store append-only events, group them into runs, calculate risk flags, and preserve compatibility with the original x402 event model.
The dashboard shows recent runs, selected timelines, approval/payment/error evidence, spend by endpoint, spend by agent, and audit export.
The Splunk HEC adapter exports a selected run as agentops:run_event records for search, alerting, and incident review.
The credibility layer is intentionally boring: public code, repeatable demo data, tests, release assets, hosted page, and a sanitized Splunk proof packet.
SDK helpers, server APIs, dashboard, seed data, architecture docs, and release assets are in the public repository.
The blocked demo run exported through HEC and appeared in Splunk as five agentops-ledger events.
NandaHack is submitted and the Devpost video host is live; Splunk Devpost and broader social launch are still pending account-gated steps.
Current next gate: submit Splunk Agentic Ops with the public hosted video, repository, architecture, Splunk proof, and investigation pack. The Splunk investigation pack adds SPL searches, a Splunk MCP tool map, and AI/MCP prompt templates for the indexed run events. The Splunk submission packet keeps the Devpost copy and proof boundary in one place.
Agent Payment Ledger is positioned as infrastructure, not a novelty demo. It gives a stock agent a practical trust, audit, and observability layer.
Fits enterprise-agent trust infrastructure across safety, integration, coordination, and production-like evaluation.
Turns every agent run into Splunk-ready operational telemetry for search, alerting, and investigation.
Google Rapid Agent and UiPath remain adaptation targets after adding platform-specific runtime integration.
One place to inspect the product, code, architecture, proof, and demo material.
Shareable evidence board with the hosted video, Splunk packet, MCP map, and open gates.
One root for status, proof links, screenshots, blockers, last verification date, and machine-readable JSON.
Source, SDK, server, dashboard, adapter, docs, and tests.
Data flow from agent SDK through ledger, export API, Splunk HEC, and Splunk AI workflows.
Sanitized evidence from a local Splunk Enterprise index run.
Reusable SPL searches, Splunk AI Assistant / MCP prompts, and splunk_run_query tool mapping for agentops:run_event data.
Devpost-ready copy, evidence links, MCP alignment, and honest pending-submission status.
Captioned 46-second demo asset and packaged release notes.
Public checklist for submissions, launch links, feedback, and confirmed prize or no-prize results.
Public review thread for missing event types, approval gates, spend evidence, and Splunk handoff ideas.
Upload metadata, Devpost checklist, and the 1280x720 agentops-ledger-video-thumbnail.png asset.
Public short product walkthrough for hackathon review.
Public static demo with dashboard screenshot and submission assets.