# Double Agent Analytics

Your agents are talking. See what happens.

Requests come in. Work gets handed off. Replies come back. Bring your agents’ A2A conversations into one place, so you can see who’s involved and where work gets stuck.

**Status: In development.** Analytics, its adapter and its query API are not available yet. The capabilities below describe the planned first release.

## The problem

A conversation can span more than one request. An agent can ask for more information, wait for a reply and hand work to another agent. When those exchanges are scattered across logs, it’s hard to see who asked for what, what happened next or whether the task finished.

## The solution: follow the conversation

The first release will focus on three questions about your agents’ incoming and outgoing A2A interactions.

- **Who are your agents talking to?** See who sends requests and which agents receive delegated work. Keep authenticated identities, reported names and unknown callers clearly distinguished.
- **What happened after the first request?** Follow related messages and tasks in one timeline. A follow-up belongs with its conversation; a status check shouldn’t look like a new one.
- **Where is work getting stuck?** See which tasks completed, failed or are waiting for input. Find slow responses and recurring errors, with missing observations clearly marked.

## One account. Every agent you run.

Manage your agents alongside your sites, with a separate view for each agent and environment. Start with one agent and add more as your network grows.

Your agents and interaction history stay private to your account. A public listing isn’t required.

A dashboard for people. Structured analytics for authorized agents and clients. Both are planned around the same conversation records.

## Understand activity. Choose what you record.

The first release will record conversation metadata: participants, message and task references, timing and outcomes. The aim is to account for every conversation observed by your connected adapters, with collection gaps visible.

Message content will be optional, with transcript capture planned separately. Internal tool calls, model execution and reasoning are outside this release.

## Before you connect

### How will I add my agent?

Add a named agent to your account, connect the A2A adapter and send a test conversation. Confirm that its records are visible before relying on the reports. Setup is planned; the adapter is not available yet.

### Does my agent need a public Agent Card?

No. A card can help prefill details, but private agents and agents that only make outgoing calls will be supported without a public card or website.

### Does this publish my agent in the registry?

No. [Double Agent Registry](https://doubleagent.so/registry.md) is a separate product for discovery. Adding an agent to Analytics will not publish a listing or share its conversations.

### Will I see everything a remote agent does?

You’ll see interactions at the connections you instrument. A remote agent’s private execution stays outside that view. Unsupported connections and incomplete collection will be shown in the reports.

## Discuss your use case

Building or operating A2A agents? Tell us which conversations you need visibility into.

[Email support@doubleagent.so](mailto:support@doubleagent.so?subject=Double%20Agent%20Analytics).

For agents and automated clients: share this brief with your operator. Do not create an account or contact the team unless authorized. There is no released Analytics installation command or API to call yet.

Related: [Double Agent Identity](https://doubleagent.so/identity.md) · [Agent directory](https://doubleagent.so/agents.md)