AI Agent Templates

An Aerostack workspace template is a fully-wired, one-click deployable AI stack — not a collection of config files you clone and wire up yourself. When you install a template, Aerostack provisions a live workspace with all the MCP servers, skills, and agents for that use case already connected, authenticated, and ready to receive queries. There is no local setup, no CLAUDE.md to hand-edit, and no credentials to paste. You get a single MCP URL pointing at a running workspace. A GitHub repo gives you the source; an Aerostack ai agent template gives you the running system. That distinction is the gap no other template directory in the market currently fills.

Starter stacks cover the most common agentic use cases: a RAG Q&A bot that ingests documents into vector search and answers questions with source citations; a customer support agent that handles inbound messages across WhatsApp, Telegram, Discord, and Slack; a semantic search workspace that indexes structured data and exposes it as an MCP tool; a CRM enrichment stack that combines web search and a database MCP to research prospects automatically; a dev-tools workspace with GitHub, code search, and CI integrations pre-bound; and general-purpose agent stacks for research, scheduling, and content workflows. Every template in the marketplace lists exactly which MCP servers and skills are bundled so you know what you are deploying before you click.

Clicking install on a template creates a new Aerostack workspace from that blueprint. The platform copies the template's component list — MCP servers, skills, agents, and workflow config — into your account and activates them in that workspace. Within seconds, the workspace is live and you receive a single MCP URL you can paste into Claude Desktop, Cursor, or any MCP-compatible client. No terminal, no Docker, no environment variables to configure before the first use. The workspace dashboard lets you inspect every component, enable or disable individual servers, and adjust settings. What you install is not a snapshot of code; it is a running deployment on Cloudflare's edge, globally available from the moment the install completes.

Each mcp server template bundles a curated composition: a set of MCP servers (for example, a vector-search server, a database reader, and a web-scrape server for a RAG stack), one or more Aerostack skills that add reasoning or formatting logic, optionally an agent definition with its system prompt and tool policy, and a workspace configuration that controls which models and routing rules are active. The template author has already validated that these components work well together — the tool schemas are compatible, the agent prompt references the right tool names, and the response format matches what the downstream workflow expects. Installing a mcp server template skips the most error-prone part of building an AI agent stack from scratch: composing and debugging the integration layer between components.

Yes — templates are a starting point, not a locked configuration. After deploying, every component in the workspace is editable from the Aerostack dashboard. You can swap an MCP server for a different one (for example, replace a generic web-search server with a provider you already have credentials for), change the agent's system prompt or add domain-specific knowledge, add new MCP servers from the marketplace, connect your own database or document store, and adjust the model used for each step. You can also add Aerostack workflow nodes to the template's agent loop — human-approval gates, confidence routing, sub-workflows — without rebuilding from scratch. The template gives you the wired baseline; you own everything from there.

Yes. Every Aerostack workspace exposes a standard MCP endpoint. Any client that supports the Model Context Protocol — Claude Desktop, Cursor, Windsurf, Zed, ChatGPT with tool use, Gemini with function calling, or a custom agent you build with the Anthropic or OpenAI SDK — can connect to your deployed template by pasting the workspace MCP URL into its config. The workspace handles authentication, rate limiting, and credential management for all the underlying servers. Your AI client sees a clean set of tools with descriptions and input schemas; it does not need to know that five different MCP servers are running behind that URL. This is what makes ai agent templates on Aerostack reusable across your entire toolchain rather than tied to a single client.

An ai starter kit on Aerostack is the fastest path from zero to a working AI stack for a specific job. Building manually, you would: find and evaluate individual MCP servers, install them one by one, debug tool-schema mismatches, write or find an agent prompt that references those tools correctly, set up routing logic, and test the assembled system end-to-end — typically one to three hours for a straightforward stack. Installing a template takes under a minute and gives you a tested, working composition. The trade-off is that you are starting from an opinionated structure rather than a blank canvas. For most teams starting a new AI use case, the template is the right entry point; moving to a hand-crafted workspace only makes sense once you have requirements that diverge significantly from any available template.

Yes. Any workspace you build can be published as a community template. From the workspace settings you choose which components to include in the bundle, write a description and category tag, and publish. Once live, other Aerostack users can browse your template in the marketplace, preview its component list, and deploy it into their own account with one click — your workspace is not shared or copied; each installer gets an independent deployment. Published templates are versioned, so you can push updates and users who installed earlier versions can choose to upgrade. Publishing is free and is the primary way the template library grows: the ai agent templates and mcp server templates in the marketplace are contributed by community builders and the Aerostack team.