---
title: "Platform capabilities"
description: "Compare models, release tasks, set agent permissions, share instructions, and learn from production work."
order: 9
---

# Platform capabilities

Orlo helps a team decide how AI should work, use that decision in production, and improve it from real results.

| What you want to do | How Orlo helps | Start here |
| --- | --- | --- |
| Choose a model for a task | Compare candidate answers on the same examples and rubric. See scores, uncertainty, cost, and failures before choosing. | [First task](/start/first-task) |
| Run a task consistently | Release a model with a task version and saved validation and routing settings. Inspect answers against that release. | [Compare and release](/console/tasks/compare-and-release) |
| Answer from your documents | Add task sources and inspect the passages retrieved for each answer. | [Sources](/console/tasks/sources) |
| Set what an agent can do | Check wrapped actions against permissions, limits, and human approvals. Inspect each run and its decisions. | [First agent](/start/first-agent) |
| Share instructions | Publish skill versions for employee AI tools and application agents. Track retrieval and declared use. | [Team skills](/start/team-skills) |
| Improve the next release | Review corrections and saved steps, add accepted examples to a dataset version, and compare again. | [Improve from production](/console/tasks/improve-from-production) |
| Review work across a team | Give people the access they need, route approvals, and inspect production and change records. | [Review work](/start/review-work) |

Applications connect through the task API, a text chat-completions endpoint, or the TypeScript and Python SDK integrations. Teams can use Orlo with an existing model gateway. [Choose an integration](/engineers).

For deployment options, see [Operate Orlo](/operators). For the public SDK source and packages, see [Open source](/open-source).