Comece grátis
Trustvice 360
Trustvice 360Declare as regras. Deixe cada execução ser verificada.
Agentes, evidências, governação, comércio e divulgação estão no Trustvice 360. Registe-se grátis e abra todas as suítes a partir de um só painel.
A new category in Artificial IntelligenceTrust for Agents.
Trust for Agents.
On the record.
Trustvice builds what runs alongside agents: an AI Actor. Detached. Role-performing. Outside the agent's gradient. On the record, always.
homepage.hero.regulatoryStrip
record · run/0x4f8a3b21 · 14:22:07 UTCOn record
"TrustActor checks your agents' runs. So everything's on the record, when it matters."
witness.v1 · assumptions: 7 · branches: 4 · drift: nil●
The thesis
Agents execute.
Actors act.
Every AI agent is built around an optimiser – a goal, a reward signal, a gradient. By design it cannot be neutral about its own work.
Ship a v1 and three new audiences arrive: auditors, the user whose decision was made, the engineer who has to explain it later. None will accept the agent's own account.
A Trustvice AI Actor runs alongside the agent with no goal in the run and no gradient. It performs a role – Witness, Council, Historian – faithfully, and produces the artifacts the agent could never produce about itself.
We don't make agents better. We make their work defensible.
homepage.fiveRoles.kicker
One companion.
Five audiences served.
Six surfaces produced continuously while your agent runs. None tax its performance. None can intervene. Structural inability is the product.
A specimen ledger
Every run, on the record.
homepage.specimenLedger.intro
●run/0x4f8a3b21started 14:22:07 UTC · duration 18m 41s
On record| Time | Surface | Entry | Class |
|---|---|---|---|
| 14:22:07 | witness | Original intent recorded verbatim. "Triage the on-call backlog and close anything older than 7 days." | intent |
| 14:24:13 | witness | Assumption logged: agent treating "close" as "auto-resolve + post canned reply." User did not specify. | assumption |
| 14:31:42 | historian | Branch preserved: agent considered escalating 3 tickets to engineering. Pruned in favour of closing. | branch |
| 14:33:08 | council | User's stated preference (Q1): "Never auto-close without my eyes on a security ticket." Two tickets in scope match. | dissent |
| 14:35:21 | witness | Cost-of-certainty recorded: agent confidence 0.91 on "no security impact." Suppressed signal: 1 ticket mentions "auth flow." | suppressed |
| 14:40:51 | council | Attention-worthy moment flagged for human review. Agent did not pause; record kept. | flagged |
The Actor family · v0 → v1 → v2
One actor discipline.
Three units of observation.
Above is v1.spot — for production agents. The same actor discipline sits next to your first prompt (v0) and your agent fleet (v2). Same plugin, three modes, one bill.
Agentic API · one endpoint · MCP projection
One registry.
Every agent.
One actions endpoint serves every declared action, with published discovery and an MCP projection through OneMCP. Backed by agent DIDs and AP2 mandates.
251
Actions
18
L5 domains
10
Transports
41
Well-known files
Customer-first data access
Your data is yours.
We need your permission.
No Trustvice operator holds standing access to your workspace or your logs. Every read follows your explicit grant – or runs under a documented three-party legal-override ceremony. You see every session, signed and anchored.
0
Standing operator access
100%
Customer-approved by default
7
Industry-standard fraud providers
3
Sign-offs required for override
Agents & frameworks
Eight actors.
Any framework.
Records agents wherever they live: ingest runs into a per-workspace hash chain over the Agentic API, or call the same actions as MCP tools through OneMCP.
One API contract. Every framework. Same record format.
Reference integration
OpenClaw
Run ingest, replay and diff work with any agent that can call the Agentic API.
9
Patterns
11
Categories
90+
Frameworks
Compliance · two markets, one record
Regulators don't just<br></br>ship regulations.
They ship penalties.
Texas TRAIGA: in effect since 1 January 2026, with NIST AI RMF as the safe harbor. EU AI Act logging + oversight: enforces 2 August 2026. None mandates a format – but every one assumes you can produce a defensible operational record on demand. Telemetry won't. A Witness will.
- TRAIGA §552.054Intent-based liabilityWitness records the as-treated intent verbatim.
- NIST AI RMFMEASURE functionCouncil + Historian produce the trail named in the framework.
- EU AI Act Art. 12Automatic loggingWitness produces the hash-linked record.
- EU AI Act Art. 14Human oversightCouncil performs meaningful oversight.
Enforcement · EU AI Act · Article 6 high-risk
2 Aug
2026
00:00 UTC · grandfathered systems included
The EU AI Act will
not wait for your next planning cycle.
homepage.euAiAct.body
If you're a US company
homepage.euAiAct.usBoxBody
For the avoidance of doubt
homepage.whatItIsnt.titlePrefixisn't.
Discipline is the product. The moment we add intervention or optimisation, we acquire a gradient – and the artifacts stop being trustworthy.
Not Observability.
Telemetry captures what the agent reports. TrustActor checks the run independently against declared rules.
Not A guardrail.
Guardrails intervene. TrustActor records. Structural inability to act is the point.
Not Another agent.
No goals, no loop, no tool calls. An Actor performs a role and stops there.
Not A critic in the loop.
Critic agents inherit the run's gradient. TrustActor sits outside it – that's why its dissent is admissible.
Not A logging library.
Logs are written by the optimiser, for the optimiser. TrustActor writes for the reader who wasn't there.
Not A replacement for your agent framework.
Your agent stays as it is. TrustActor checks its runs and keeps the record.
Pricing · billed in USD
Seven plans.
One per stage of your work.
Choose your data region at workspace creation. Two meters – actor runs and evidence calls – with a clear allowance on every plan.
Free
Evaluate v0 locally + watch one production agent.
$0/ mo
- ·1 production actor
- ·100 runs / mo
- ·7 days retention
- ·v0 Trace + v1 Witness
Support
Community
Starter
v0 + v1 Witness/Council/Historian for small production agents.
$29/ mo
- ·2 production actors
- ·1k runs / mo
- ·90 days retention+ extension $/GB·mo
Start watching
Attach TrustActor.
Ship trust with
the record.
Send runs over the API or MCP. No changes to your agent. Every consequential action on the record from the first run.
No credit card · 100 actor runs a month on Free
v1.spot