# AI run limits

Bound repeated AI processing and function use while allowing the agent to produce a final answer.

*Release 3.3 · developer*

**Settings › AI › Features** · **Settings › AI › Accounts**

## What it is

An AI agent may need several model requests to answer a question. It can request functions, read their results and ask the model what to do next. Run limits restrict this repeated work and further function use.

## Who uses it and why

Developers use these limits when designing integrations that need several searches or actions. Ask Powerfront to adjust account allowances when a supported workflow needs more processing time.

## Where to find it

Feature options are under **Settings › AI › Features**. You need the **Developer** role, or **Settings** with **AI** allowed under **Settings Permissions**.

Powerfront manages **Max Function Count** and **Max Function Seconds** under **Settings › AI › Accounts**. Ask Powerfront to change these account settings for you.

## How it works

1. INSIDE prepares the selected feature's instructions, conversation context and available functions.
2. The model generates a response or requests functions.
3. INSIDE executes available function requests and returns their results to the model.
4. Processing repeats until the agent replies, hands over control, is blocked, fails or is cancelled.

INSIDE checks the generation-round and elapsed-time allowances between rounds. Reaching an allowance disables further function use and lets the model produce a final answer. Additional run-specific controls can restrict total function calls and token use.

Integrations can observe processing stages, model requests, function execution and outgoing messages. Some configured integrations can change or block processing before the next action. These integrations are arranged through Powerfront.

Ending a conversation cancels its active AI work. A replacement request can cancel an older response. With managed context, new input can instead join the active run.

## Configure it

Choose the feature's **Account** and set **Max Tokens** if you need a feature-specific generation allowance. A value of **0** uses the account default.

Ask Powerfront to review **Max Function Count** and **Max Function Seconds** with representative requests. The normal defaults are **3** and **30 seconds**. The account editor accepts counts from **1** to **10** and time allowances from **1** to **60 seconds**.

Managed context uses **Maximum run time (seconds)** in place of the normal function-time allowance. Review that setup separately before extending long-running work.

### Settings

| Setting | Where | Applies to | Default | Who can change it | What it does |
|---|---|---|---|---|---|
| Max Function Count | Settings › AI › Accounts | website | 3 | Powerfront, on request | Sets the normal number of generation rounds in which the agent may use functions. |
| Max Function Seconds | Settings › AI › Accounts | website | 30 seconds | Powerfront, on request | Sets the normal elapsed-time allowance for further function use when managed context is not active. |
| Max Tokens | Settings › AI › Features | website | 0, use the account default | Settings role, Developer role | Overrides the account's generation token allowance when a positive value is entered. |

### Permissions

| Permission | Lets you |
|---|---|
| Settings | Configure feature-level options when AI is allowed under Settings Permissions. |
| AI | Allows AI configuration within Settings Permissions. |
| Developer | Configure feature-level options. |

## Limits and good to know

- **Max Function Count** counts generation rounds, despite its label. One round may request more than one function.
- Reaching a loop or time allowance stops further tool use. It does not guarantee the whole response finishes at that exact second.
- A provider request or external function may have its own timeout.
- These controls describe standard text-agent execution. Email Autopilot and Voice Autopilot have separate handling.
- A skill cannot increase the originating feature's execution allowances.

## Depends on and works with

### This feature uses

- **Configured in** [Roles and permissions](ADM-008.md): Roles and AI permissions control feature configuration access.
- **Configured in** [AI feature settings and change history](AI-001.md): The feature selects the account and feature token allowance.
- **Configured in** [Context Compiler for AI conversations](AI-015.md): Managed context uses its maximum run time instead of the normal function-seconds allowance.
- **Configured in** [AI accounts and shared defaults](AIP-001.md): Powerfront manages the account's normal execution limits.
- **Works better with** [AI functions and required inputs](AI-016.md): The function catalogue supplies the operations used between generations.
- **Works better with** [AI guardrail routing and fallback actions](AI-025.md): Guardrails can block or change processing before an answer is delivered.
- **Works better with** [AI reply streaming settings](AI-037.md): Streaming can display a running reply progressively and control interrupted drafts.
- **Works better with** [Virtual Assistant tasks started by Autopilot](ENG-027.md): A workflow handover can end AI handling of the current request.
- **Related** [AI agent service connections](AI-024.md): The selected execution service determines whether built-in AI flow limits apply.

### Used by

- **Required by** [Context Compiler for AI conversations](AI-015.md): Managed continuation operates within an AI run and its remaining execution allowances.
- **Required by** [Reusable AI skills and instructions](AI-020.md): Skills operate inside the feature's run and cannot enlarge its execution allowances.
- **Required by** [Customer goals and AI progress tracking](AI-021.md): AI execution creates and updates goals through authorised operations.
- **Required by** [AI conversation plans and task progress](AI-022.md): AI execution manages plans and their task state.
- **Required by** [Specialist sub-agents for AI tasks](AI-023.md): Specialist work runs within execution limits.
- **Required by** [Conversation history tools for AI](AI-038.md): AI flow execution provides the built-in history tools and session context.
- **Required by** [Images and attachments for AI agents](AI-039.md): AI execution prepares image input and supplies image viewing tools.
- **Configures** [AI functions and required inputs](AI-016.md): Run allowances restrict repeated function execution.
- **Configures** [AI knowledge and web search](AI-017.md): Searches and page reads consume the agent's execution allowance.
- **Configures** [AI integrations and automation tools](AI-018.md): External operations consume the owning run's function allowance.
- **Configures** [AI assistance activity](AI-033.md): Execution limits bound repeated generation and tool work.
- **Configures** [AI platform execution limits](AIP-024.md): The feature's Max Tokens and its account's function allowances control standard text-agent runs.
- **Enhances** [AI agent service connections](AI-024.md): AiFlow is the built-in execution option and an optional fallback.
- **Related** [Virtual Assistant tasks started by Autopilot](ENG-027.md): Accepted task handoff ends the current AI response and skips later functions in the same batch.

## FAQ and troubleshooting

### Why did the AI answer without completing another lookup?

It may have exhausted its function-use allowance. Review the request and account settings with Powerfront before increasing the limits.

### Does Max Function Count of 3 mean exactly three function calls?

No. It limits generation rounds that may use functions. A round can contain several calls.

### Why did the response take longer than Max Function Seconds?

The elapsed-time check prevents further function use between rounds. An in-progress request and the final answer can take additional time.

### Can I change the account limits myself?

Powerfront manages the account screen. Ask Powerfront to change the limits and explain which requests need a larger allowance.