---
title: Get started with 0.2.4
description: Configure models, add and share resources, and connect AI employees to business systems.
lastVerified: 2026-10-22
---

# 1.2 Get started with 0.2.4

This guide follows the actual Community 0.2.4 workflow. For an existing installation, follow [Upgrade and rollback](https://docs.teloa.ai/v/0.2.6/markdown/en/deploy/backup.md) to preserve its data and installation identity. You do not need to recreate employees or migrate conversations.

## 1. Configure your first chat model

Open **Settings > Models**, add a provider, enter its key, and use the **Save** button on that provider’s own card. When this is the first usable provider and the previous default has no credentials, blank conversations use the new configuration. An existing valid default or an explicit model choice is preserved. Check the model in the composer, then send a simple question. Group routing uses the default model.

Chat and speech output store credentials separately. “Credentials saved” on the voice page does not mean a chat model is configured. See [Models and credentials](https://docs.teloa.ai/v/0.2.6/markdown/en/start/models.md) for setup and troubleshooting.

[![The workspace with a goal draft and ready tasks. The example input has not been sent.](https://docs.teloa.ai/v/0.2.6/assets/screenshots/en/workbench-024.18aefe7ecd43.png)](https://docs.teloa.ai/v/0.2.6/assets/screenshots/en/workbench-024.18aefe7ecd43.png)

*0.2.4 actual interface with example data.*

## 2. Create an employee from a short description

Open **Employees > Add AI employee**. Describe the responsibility, source material, deliverable, and actions that need your confirmation. For example:

```text
Draft product update notes from the material I provide. Cite sources and list open questions.
You may organize and revise drafts. Ask me before publishing, sending, or expanding the source scope.
```

Continue to the original conversation, review the prepared input, and send it. Once the model produces a configuration draft, review its responsibility, scope and resources, then explicitly create the employee. Its profile provides capabilities, memory and runtime settings. Pause the employee before editing settings that require it, save, and resume. See [Employees](https://docs.teloa.ai/v/0.2.6/markdown/en/guides/colleagues.md).

[![Describe responsibilities, sources and boundaries in Markdown. The shown draft has not been continued or submitted.](https://docs.teloa.ai/v/0.2.6/assets/screenshots/en/employee-create-024.d5c8f0263fea.png)](https://docs.teloa.ai/v/0.2.6/assets/screenshots/en/employee-create-024.d5c8f0263fea.png)

*0.2.4 actual interface with example data.*

## 3. Add or share four resource types

The **Marketplace** has **AI employees, AI teams, Apps and Skills**. “All” combines these four categories. “My apps and skills” manages installed methods and connections; “Added” in the marketplace retains resources you have added.

Open details, check contents, provenance and dependencies, then confirm adding. A team template may include several employees, skills, materials and routines. Check employee settings and account connections separately, and enable schedules explicitly. Adding a template does not start work.

Open sharing from an employee, team, app or skill. Review the files and licenses before exporting an offline template. Online sharing opens the Teloa marketplace account page: review the complete public contents and explicitly publish, then return to the workspace to copy the link. The publisher can revoke the link; revocation does not delete files already downloaded by others. Recipients preview the package, add their own copy, and configure accounts and permissions. See [Adding and sharing resources](https://docs.teloa.ai/v/0.2.6/markdown/en/guides/market.md).

[![The marketplace categories are AI employees, AI teams, Apps and Skills, with entries for added content and app management.](https://docs.teloa.ai/v/0.2.6/assets/screenshots/en/market-024.324853296f4d.png)](https://docs.teloa.ai/v/0.2.6/assets/screenshots/en/market-024.324853296f4d.png)

*0.2.4 actual interface with example data.*

## 4. Prepare local models and voice

Local chat and retrieval settings are together under **Settings > Local models**. The Library shows retrieval status and an entry to these settings. Chat models generate answers; retrieval models build a material index. Prepare them separately.

Retrieval offers Qwen3 Embedding 0.6B or Google EmbeddingGemma 2. The latter requires Ollama installed and running on the Teloa host. The former downloads the components listed on its preparation page. Check size, source and runtime requirements before confirming a download. When ready, select the model and add materials to the index. “Enabled” alone does not mean preparation is complete.

In **Settings > Voice service**, prepare local SenseVoice recognition and configure MiniMax speech output separately. Downloads support cancellation and failure retry. First use requires microphone permission. Review transcription before sending, and use the preview to check speech output. See [Model and voice setup](https://docs.teloa.ai/v/0.2.6/markdown/en/start/models.md).

[![Local model and retrieval settings. Ollama is not connected and retrieval is not enabled in this example.](https://docs.teloa.ai/v/0.2.6/assets/screenshots/en/local-models-024.8ae5acc6dfb0.png)](https://docs.teloa.ai/v/0.2.6/assets/screenshots/en/local-models-024.8ae5acc6dfb0.png)

*0.2.4 actual interface with example data.*

## 5. Enable browser and computer operation

Open **Settings > Connections and environment**. Browser Use requires `@teloa/native-browser` and its Browser Use and Playwright providers. Computer Use uses `@teloa/native-computer`, Computer Use and Cua. Follow the page to prepare an installed browser or system permissions. When a restart is required, wait for work to finish, restart, and open a new conversation.

Once the browser bundle is enabled, **Start > Browser** in the right sidebar is for manual browsing. The assistant’s Browser Use operates isolated headless pages without asking for a PID. CUA operates the physical desktop; specify the target application. Enabling a package does not grant every employee its tools. See [Native capabilities](https://docs.teloa.ai/v/0.2.6/markdown/en/guides/native-capabilities.md).

[![Inspect and selectively enable native capabilities in Connections and environment. Enablement does not grant employee permissions.](https://docs.teloa.ai/v/0.2.6/assets/screenshots/en/extensions-024.58d3171491f5.png)](https://docs.teloa.ai/v/0.2.6/assets/screenshots/en/extensions-024.58d3171491f5.png)

*0.2.4 actual interface with example data.*

## 6. Call employees through an SDK, API or MCP

In **Settings > Employee integrations**, select an employee and business scope. Optionally allow starting, continuing or stopping work, then create integration credentials. The key appears once. Store it on your business system’s server, never in chat or a browser frontend.

- HTTP: read the host’s `/api/v1/openapi.json` contract.
- TypeScript / JavaScript: install `@teloa/sdk` and create `TeloaClient` with the integration key.
- MCP: connect to the host’s `/mcp` using the same integration Bearer Key.
- Asynchronous updates: fix the callback URL when creating the integration, verify CloudEvents signatures, and persist event IDs for deduplication.

Persist an idempotency key before starting work. Keep the returned work ID, query status, and read the fixed result. If a response is unconfirmed, reconcile using the original identity to avoid duplicate execution. The employee’s BYOK, tool permissions and work approvals still apply. See [Employee integrations](https://docs.teloa.ai/v/0.2.6/markdown/en/develop/employee-api.md) for complete examples.

[![Scope and permission controls before creating integration credentials. No key or work has been created.](https://docs.teloa.ai/v/0.2.6/assets/screenshots/en/employee-api-024.971227cf2aa8.png)](https://docs.teloa.ai/v/0.2.6/assets/screenshots/en/employee-api-024.971227cf2aa8.png)

*0.2.4 actual interface with example data.*

## 7. Inspect context and usage

Open a conversation’s **Context** tab to inspect messages, instructions, tools and the usage statistics supplied by the runtime. Ordinary conversations discover and load platform tools as needed; employee tool access still depends on authorization. A short visible chat does not mean the model receives only those messages: responsibilities, materials and runtime instructions can also occupy context. Use the actual breakdown to decide whether to shorten material or start another topic. Provider billing remains authoritative.

## Next step

Complete a small task without external sending, check actual items in **Needs your attention**, and review results and sources in the original task. A successful conversation, a completed tool call and business acceptance are different outcomes. Check the current permission mode before changing it: Auto review uses model review; explicit Full access does not add that review. Neither expands an employee’s granted tools or business scope. Keep the default permissions for routine work. See [Permissions](https://docs.teloa.ai/v/0.2.6/markdown/en/develop/permissions.md).

To build and distribute your own resources, follow the [four authoring paths](https://docs.teloa.ai/v/0.2.6/markdown/en/develop/overview.md) and [sharing guide](https://docs.teloa.ai/v/0.2.6/markdown/en/guides/sharing.md), including downloadable examples and official review.