Skip to content

Find models, tools and integrations in the Marketplace

Updated

The Marketplace is where you find the models, tools, integrations, agents and skills your agents and Omni can use. It sits alongside the aixplain OS the way an app store sits alongside an operating system: the OS runs your agents, and the Marketplace supplies what they work with, all reached with one aixplain API key. This article explains what you can find there, how to search and filter, how pricing works, how to add an asset to an agent or to Omni, how to bring your own and what you can use on your own servers.

For a tour of the rest of the app, see Use the aixplain OS.

What's in the Marketplace

The Marketplace lists more than 900 assets, such as models, tools and agents, and more than 600 integrations. Each asset has a card and a page with its description, its price in credits and how to call it.

Asset typeWhat it is
ModelsLLMs and other models, such as speech recognition, translation, image and video generation, that you run directly or give to an agent
ToolsHosted tools ready to run with no connect step, such as web search, a Python sandbox, a file manager and Marketplace search
IntegrationsApps you connect with your own account, such as Slack, Gmail or Salesforce, plus databases and MCP servers
AgentsAgents that other teams built, tested, deployed and published for you to use
SkillsReusable instructions, scripts and reference files that an agent loads when a task needs them
GuardrailsReady-made checks you attach to an agent as inspectors, such as prompt-attack detection or sensitive-information filters

Every model and tool in the Marketplace is also a hosted MCP server, so you can call it from any MCP client with your API key. See MCP servers in the Agents SDK docs.

Models and providers

The Marketplace carries frontier models, open models and Arabic-first models from many providers, so you can pick or swap the model behind an agent without a separate account with each provider. Families you'll find include:

  • Frontier models: Claude Opus, GPT and Gemini.
  • Open models: Qwen, DeepSeek, Kimi and GLM.
  • Arabic-first models: ALLaM and Jais.
  • Speech, image and video: Whisper and Deepgram Nova for transcription, Gemini Flash Image for images and Veo for video.
  • Athena: aixplain's meta-model, which picks the right model for each request.

The same model can be served by more than one host, such as the provider itself or Bedrock. A model's page shows its host and its data retention, and lists its actions and functions.

Tools and integrations

Tools include web search from Tavily, Exa and Brave Search, Google Places search, Google Cloud Translation, a Python sandbox and Marketplace search. Integrations include Slack, Gmail, Google Drive, Google Calendar, Outlook, Salesforce, HubSpot, Jira, Notion, GitHub and Airtable, plus PostgreSQL and SQLite databases and any MCP server reachable over HTTP, including servers that sign in with OAuth.

Search and filter

  1. Go to app.aixplain.com/marketplace.
  2. Type what you need in the search box, such as "transcription" or "Slack". You can also search by an asset's ID.
  3. Narrow the results with the filters.
  4. Open a card to see the asset's page.

Search covers models, tools, integrations, skills and files, and results come back in order of relevance. Your team's own assets show up in results too.

FilterWhat it does
CategoryShows one kind of asset, such as LLMs
DeveloperShows assets built by one developer
Data retention (ZDR)On the Models page, shows only zero data retention models, which keep nothing you send

Each asset also lists where it is hosted. In the Agents SDK and the Marketplace search tool you can filter by host and supplier as well, which helps when your data has to stay with a given provider or in a given place.

Prices and credits

Each card shows the asset's price in credits, and one credit is one US dollar. Providers set their own prices:

  • Models: usually per token. A model's page shows its input, output and cached-input prices, and cached prompt tokens are billed at the cached-input price. Other models are priced per unit, such as audio minutes, images or seconds of video.
  • Tools: usually per call.
  • Integrations: not billed themselves. The tools you create from an integration are billed as tools.
  • Agent runs: the sum of every model and tool call in the run, plus 20% for orchestration, inspectors, memory and tracing. Calling a model or tool directly has no aixplain fee.

For example, a model at $10 per million input tokens gets you about 100,000 input tokens for one credit. Every run's cost shows in its trace and on the usage page. Local models and tools in aixplain Desktop don't use credits. For plans, top-ups and budgets, see Your account, plans and billing.

Add an asset to an agent or to Omni

Add it to an agent

In Studio, open your agent and add the models, tools, integrations, skills and guardrails it needs from the Marketplace. You can also change the model an agent reasons with, and the agent keeps its owner, policies and history.

With the Agents SDK, get the asset by its path and pass it to the agent: a model in llm, tools in tools, skills in skills and guardrails in inspectors. For each tool, set the actions the agent may use, so it sees only what it needs. For code, see the Agents SDK docs, and for building in the app, see Build and deploy agents.

Use it with Omni

You don't need to add tools to Omni by hand. Omni discovers the tools a task needs in the Marketplace and asks you to connect any you haven't yet. When you're signed in, it also sees your team's agents, models, connections and tools, so it can pick the right one. To give the workspace agent behind chat its own tools, integrations and models, set it up from the agent library. See Get started with Omni.

Connect an integration

An integration acts in an app with your account, so it needs a connection before an agent can use it.

  1. Open the integration's page in the Marketplace, or add it to an agent.
  2. Sign in to the app with OAuth, or enter a token or API key, depending on what the app supports.
  3. Choose which actions to connect.

The result is a tool in your workspace that any of your agents can use. Credentials for integrations are stored by the integration provider, not by aixplain.

Bring your own models and tools

Your own assets sit next to the Marketplace's, private in your workspace's Library. Only your workspace sees them unless you share them.

  • Your own tools: write a Python function and save it as a script tool, or connect your own PostgreSQL or SQLite database.
  • Your own APIs and apps: connect any MCP server reachable over HTTP as a tool.
  • Your own skills: upload a folder with a SKILL.md file, then attach the skill to any agent.
  • Your own models: connect your own model servers, or bring your own keys for a provider.

To build these with code, see the Agents SDK docs. For who in your workspace can see and change them, see Manage your workspace.

Share or publish an agent

An agent you build can stay private in your Library, be shared with the workspaces you choose or be published to the Marketplace for others to find and use. Skills published by other teams appear on the Marketplace and in the agent builder's picker, and you can download a skill as a zip from its page. On the Scale plan, you can also share agents, skills and artifacts with your team. See Manage your workspace.

Use the Marketplace on your own servers

With Enterprise, the aixplain OS runs on your own servers, in your VPC or air-gapped. There you can use any model in the Marketplace, open-source models you host or your own, plus your systems connected as tools.

  • Hosted on your servers: models and tools you host use no credits.
  • Cloud models and tools: agents can run on your servers and still call Marketplace assets on cloud. You add usage credits only for those.
  • Air-gapped: nothing calls outside your network, so agents use only the models and tools you host.

See Run the aixplain OS on-prem.

Privacy and data

When an agent calls a cloud model, the request goes to that model's provider. To keep a provider from retaining what you send, pick a model marked zero data retention (ZDR). We never sell your data or train on it. See Privacy and data and Security.

Frequently asked questions

Do I need an account to browse the Marketplace?

No. You can browse the Marketplace as a guest. Adding an asset, connecting an integration or running anything asks you to sign in. Signing up is free and new accounts start with free credits.

Do I need my own account with each model provider?

No. Every model and tool in the Marketplace runs with your aixplain API key and is paid from your workspace's credits. You can still bring your own keys.

Where do I find an asset's price?

On its card and its page, in credits. Model pages show the input, output and cached-input prices.

How do I find models that keep nothing I send?

Use the Data retention (ZDR) filter on the Models page. Every model's page also shows its data retention.

Why does an integration ask me to sign in?

An integration acts in an app with your account, so you connect it once with OAuth, a token or an API key. Hosted tools such as web search need no connect step.

Can I use Marketplace assets from Claude Code, Cursor or another MCP client?

Yes. Every model and tool is a hosted MCP server that any MCP client can call with your API key. The MCP servers page in the Agents SDK docs shows how to add one.

Can I search the Marketplace from an agent?

Yes. Give the agent the Marketplace search tool. It finds agents, models, tools and integrations and reads an asset's price, host and status. Its run actions spend credits, so allow them only when the agent should run what it finds.

Why can't I see an asset my teammate made?

Assets are private to the workspace that owns them. Check that you're in the right workspace, or ask your teammate to share the asset. See Manage your workspace.

Get help

  • Learn the SDK and API: read the Agents SDK docs.
  • See what changed: read the release notes for new models, tools and Marketplace updates.
  • Talk to us: email care@aixplain.com for help with an asset, a bug or a privacy request.

Still need help?

Report a bug on GitHub or write to us, and include your version and operating system.