Running and Monitoring
=======================

Once a Neo Code Agent is published, you run it and observe it from the tabs across the top of the agent. Each tab is separately permissioned — see :doc:`permissions`.

Running the Agent
------------------

**Through NeoGPT.** The agent appears in the NeoGPT agent picker, where you can hold a conversation with it.

**Through the run API.** The **API Access** tab supplies ready-made request samples. Either way, the platform receives the call and passes it to your service as described in :doc:`service_contract`.

API Access
-----------

The **Run** tab provides **Python**, **JavaScript**, and **cURL** samples with a copy button. The endpoint is:

.. code-block:: text

   POST {AGENTIC_FLOW_URL}/api/v1/run/{agentId}

The **Parameters** panel covers ``session_id``, ``output_component``, and ``tweaks``.

.. note::
   ``tweaks`` is the parameter to reach for with a Neo Code Agent: whatever you pass there arrives at your service as the ``config`` object in the request body. Use it to hand structured settings to your own code. See :doc:`service_contract`.

Edit the sample to match the input your service expects before using it.

Conversations
--------------

The **Conversations** tab lists conversation records for the agent.

.. list-table::
   :widths: 26 74
   :header-rows: 1

   * - Column
     - Detail
   * - **Conversation ID**
     - The session identifier — the same ``session_id`` your service receives.
   * - **Status**
     - **Completed** (green) or **In Progress** (yellow).
   * - **Duration**
     - Length in minutes, or ``-`` while the conversation is still open.
   * - **Created At**
     - When the conversation started.
   * - **Actions**
     - **Mark as Complete** for an open conversation, and **Delete**.

Search covers **Conversation ID** and **User ID**. Opening a row shows the conversation's identifier, status, duration, message count, and creation time, along with **KPI Results** and **Session Metadata**.

Analytics
----------

**Analytics** uses the same dashboard as Agentic Flow agents.

Six summary cards head the page: **Total Cost**, **Total Tokens**, **Total Conversations**, **Avg Latency (s)**, **Input Tokens**, and **Output Tokens**.

Below them, **Agent Runs** lists individual executions with **Run ID**, **Session ID**, **Root Agent**, **Start Time**, **Own Cost**, **Own Tokens**, and a **View Traces** action that opens the traces for that run. Three tables follow:

.. list-table::
   :widths: 34 66
   :header-rows: 1

   * - Table
     - Columns
   * - **Sessions by User**
     - Email, Sessions, Total Cost, Total Tokens
   * - **User Level Cost and Token**
     - Email Id, Total Cost, Total Tokens
   * - **Model Cost and Token Usage**
     - Model Name, Total Cost, Total Tokens

Cost and token figures cover model usage the platform performs. Models your own service calls with its own credentials are not counted here; latency and conversation counts reflect every request Neo passes through.

Traces
-------

**Traces** lists **Trace ID**, **Conversation ID**, **Name**, and **Created At**, filterable by **Created on** and **Session** and searchable by **Name**. With dataset permissions, a row action adds a trace to a dataset for evaluation.

Traces cover the requests Neo passes to your service.

Channels
---------

The **Channels** tab offers the same channel catalogue as Agentic Flow agents. Publish the agent before attaching a channel. For channel setup, see :doc:`/13_channels/index`.

Where Else the Agent Appears
-----------------------------

On **Home**, a Neo Code Agent sits in the same grid as every other agent, and the **Type** filter includes a **Remote Agent** option for narrowing the list. Each card shows the agent's current status — whether it is deploying, deployed, disabled, being edited, or awaiting approval.

Platform-wide monitoring in the :doc:`Command Center </11_command_center/index>` covers Neo Code Agents alongside every other agent type.