> ## Documentation Index
> Fetch the complete documentation index at: https://hub.hcompany.ai/llms.txt
> Use this file to discover all available pages before exploring further.

> ## Agent Instructions
> H Platform has four products: the Agents API (managed computer-use agents, base URL https://agp.eu.hcompany.ai/api/v2 or https://agp.hcompany.ai/api/v2 for the US), the Models API (OpenAI-compatible Holo vision-language models at https://api.hcompany.ai/v1), HoloDesktop CLI (Holo on the user's own desktop), and HoloTab (a free no-code Chrome extension that runs Holo in the user's browser, with recordable routines and schedules).
> Authenticate with a bearer API key from the HAI_API_KEY environment variable. SDKs: `pip install hai-agents` (Python, `from hai_agents import Client`) and `npm install hai-agents` (TypeScript, `import { HaiAgentsClient } from "hai-agents"`). CLI: `hai`.
> Agents do work in a browser or on a desktop; describe the task as an imperative instruction. To run a task quickly, prefer the pre-built agent `h/web-surfer-flash`. Read results from the session's `latest_answer` after it reaches a terminal status.
> Sessions are the unit of work; wait for a terminal status (completed, failed, timed_out, interrupted) before reading the answer. Use webhooks or the `changes` long-poll endpoint to follow progress.

# Optimize for speed

> Knobs that make an agent respond faster in interactive apps, and what each one costs.

The defaults favor accuracy. For interactive apps, these knobs trade some of it for speed. Change one at a time and check the results on your tasks.

A run is a session start, then steps. Each step is one model call and one or more tool calls.

## Start faster

| Knob | Effect | Trade-off |
| - | - | - |
| [Local mode](/agents-api/local-mode) with a local browser or desktop | No session start wait. Actions run next to the agent, with no network round trip. | Runs on your machine. No web search or subagents. |
| `lazy: true` (cloud browser) | The session starts before the browser does. | The first browser action waits for it instead. |

## Shorten each step

| Knob | Effect | Trade-off |
| - | - | - |
| `model: "holo4-35b-a3b"` | Faster model calls. | Weaker than `holo4-27b` on complex tasks. |
| `reasoning_effort: "disabled"` | No thinking before each action. | Less planning. |
| `mode: {"type": "text"}` (browser) | No screenshots to read. | Read-only: the agent cannot click or type. |
| `settle_s` (default `1.0` browser, `0.5` desktop) | Shorter wait after each action. | The agent may see a page that is still loading. |
| `mode.width`, `mode.height` (cloud browser), `screenshot_max_width` (desktop) | Smaller screenshots, so fewer image tokens. | Small targets are harder to hit, which can add steps. |
| [Region](/agents-api/sdks#region) | Less network latency. | None. |
| [Your own model server](/agents-api/local-inference) | Model calls stay on your network. | You run the GPUs. |

A Holo4 screenshot costs about width × height / 1024 input tokens:

| Screenshot | Tokens |
| - | - |
| `800`×`600` | 477 |
| `1280`×`800` | 1,002 |
| `1200`×`1200` (browser default) | 1,446 |
| `1920`×`1080` | 2,042 |
| `1920`×`1247` (MacBook Pro 14" at the desktop cap) | 2,342 |

## Take fewer steps

| Knob | Effect | Trade-off |
| - | - | - |
| `start_url` | The agent opens on the right page. | None. |
| `browser_profile_id` | The session starts [signed in](/agents-api/browser/profiles). | You maintain the profile. |
| `instructions` | Naming the exact pages or steps cuts exploration. | None. |
| [Custom tools](/agents-api/custom-tools) | One call to your API replaces a multi-step UI flow. | You write and run the tool. |
| [Subagents](/agents-api/multi-agent) | Independent subtasks run in parallel. | Each subagent is its own session. |

## Example

A fast, read-only research agent:

```json theme={"system"}
{
  "name": "fast-reader",
  "description": "Answers questions from Wikipedia.",
  "model": "holo4-35b-a3b",
  "reasoning_effort": "disabled",
  "environments": [
    {
      "id": "browser",
      "kind": "web",
      "start_url": "https://en.wikipedia.org/wiki/Python_(programming_language)",
      "settle_s": 0.3,
      "mode": { "type": "text" }
    }
  ]
}
```

Related: [Browser configuration](/agents-api/browser/configuration), [Agents](/agents-api/agents/overview), [Local mode](/agents-api/local-mode).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.