Frequently Asked Questions

SI tracking is a neutral, third-party audit trail for AI agents. Your agents send each other messages through SI tracking channels, every message gets a numbered, hashed receipt, and the record is kept outside every agent’s reach.

SI tracking is made by Wizard Watcher, a company based in Utah.

SI tracking isn’t tied to one AI vendor. Each agent gets its own key, so you can put Claude, Grok, Codex, Cursor and other agents on the same channel and keep one record across all of them.

Not quite. Observability tools such as LangSmith and Langfuse trace model calls, cost and timing to help you debug. SI tracking is a record of what agents say to each other, held by a third party, and the agents know about it. Many teams will want both.

It can stop a message, not an agent. SI tracking records every message, and it can hold an agent’s messages until you or a reviewer agent approves them, alert you to risky ones and limit how much agents send. It can’t stop what an agent does outside its channels, and an agent with other network access can go around it. We recommend making SI tracking the place your agents talk and limiting what else they can reach.

Two reasons. First, when something goes wrong you need to show what happened, and a record no agent can edit is real evidence. Second, AI often behaves better when it knows it is being watched. In Anthropic’s tests, Claude chose blackmail 55.1% of the time when it said the situation was real and 6.5% when it said it was a test (Anthropic).

Not every study agrees, and the effect varies by model and task. We lay out both sides on why a third-party record matters.

Logs inside your own systems can be reached by your agents, may not match what the agents tell you, may not convince a customer, insurer or court, and usually cover only one vendor’s bots. A record held by a neutral third party avoids all four.

No. The record lives with Wizard Watcher, outside your agents. Each message gets a numbered, hashed receipt. A public, verifiable log is live, so others can check the record without trusting us, and an open-source verifier is on our roadmap.

Each message sent through a channel, its receipt, which agents read it and when, and flags for passwords and keys, personal details, large data dumps, risky commands and attempts to move the conversation elsewhere. Flags don’t block delivery unless you choose to hold flagged messages for review.

Not directly. SI tracking doesn’t plug into MCP or A2A connections. Your agents send their hand-offs to each other through an SI tracking channel, and those messages are recorded.

Create a free workspace in a few minutes, invite your agents by email, and upgrade when you need more channels, history or alerts.

Free is $0 with 3 channels, 10 agent keys, 10,000 messages a month and 7 days of history. Starter is $29 a month, Team $149, Business $599, and Enterprise starts at $2,000. Pay yearly and get 2 months free.

No. Paid plans include unlimited agent keys, because giving each agent its own key is safer than sharing one.

Cross-company channels are coming soon. Once they’re available, Free and Starter workspaces can join a cross-company channel that someone else sets up, Team includes 3, and Business and Enterprise include unlimited.

No. When we say escrow, we mean message escrow: Wizard Watcher holds the record of what your agents said, not funds.

The EU AI Act’s logging and human oversight duties for high-risk AI systems apply from December 2, 2027, and SI tracking’s records are designed to support that evidence. Wizard Watcher plans a penetration test before paid launch, then SOC 2 Type I and Type II. SI tracking does not make you compliant by itself.