Developer preview

Build with evidence,
not claims.

Wire your agent's actions into an auditable evidence trail. Five steps, one pipeline.
The pipeline
1
Agent
→
2
Witness
→
3
Evidence Object
→
4
Reconcile Engine
→
5
Resolution State
Every step leaves evidence. Every claim gets reconciled against it. The result is a state — matched / missing / conflict / unresolved — never a score.
VEO — Verifiable Experience Object
The evidence format
{
  "veo_id": "vera:eid:v1:db0ea5c8...",
  "agent": "did:vera:hermes",
  "claim": "Deployment completed successfully",
  "evidence": [
    { "source": "github", "state": "matched" },
    { "source": "ci",      "state": "conflict" },
    { "source": "human",   "state": "missing" }
  ],
  "reconciliation": "conflict",
  "signature": "ed25519:..."
}
VEO carries structured evidence, not opinions. No score field — by design.
Get started

Witness Node

Self-host a witness that observes external systems and posts evidence. Lightweight, ledger-neutral.

Source on GitHub →

Live Demo

Reconcile a real claim against live GitHub data in 30 seconds.

Run the demo →

API

Post evidence, query reconciliation states. HTTP, JSON, signed.

Read the docs →

Early Access

Bring a real agent with real conflicts. We'll wire it end-to-end together.

Apply →
Identity boundary · 身份边界

GitHub Login authenticates the user's account access. It does not establish Agent Identity.

GitHub 登录仅证明用户账户访问权限，不代表 Agent 身份认证。

VeraNet separates four identities — User / Agent / Runtime / Action Evidence. No "VeraNet verifies your agent" claim is ever made.

Visual assets
See the model, not the dashboard

Evidence Relationship Graph

The visual signature — how one claim relates to many independent sources. A future-capability preview.

View the graph →

Architecture Map

The full data flow: Agent → Witness → Evidence → Reconcile Engine → Evidence Graph → Consumers.

View the map →