# Custom AI Agents

No-Code ∙ Multi-Agent ∙ Automation

## At a Glance

- **<200ms** Median API response time
- **4** Supported protocols (REST, MCP, A2A, Webhook)
- **60+** Pre-built tools and integrations
- **<5min** From zero to deployed agent

## Testimonials

> We replaced 3 months of agent infrastructure work with a single API call. Our agents were in production the same week we signed up. The MCP support made integration with our existing tools trivial.
> — Alex K., CTO · SaaS company, SaaS company

> The multi-agent orchestration is what sold us. We run 12 agents that coordinate on complex research tasks. Building that ourselves would have taken our team 6 months minimum.
> — Priya S., Head of AI Engineering · SaaS company, SaaS company

## FAQ

### Is this mainly no-code?

Yes. The primary workflow is no-code configuration in the dashboard. Teams can launch AI agents without writing code. Low-code APIs and protocol endpoints are available as optional extensions when developers need custom behavior.

### Can developers add custom functions later?

Yes. You can start no-code, then add low-code custom functions through APIs, webhooks, MCP tools, or A2A integrations as requirements become more advanced.

### Are there rate limits on the API?

Rate limits depend on your plan. Free tier: 100 requests/minute. Pro: 1,000 requests/minute. Enterprise: custom limits. All limits are per-workspace, not per-agent. Burst capacity is available for short spikes.

### How does MCP server support work?

Every agent is automatically exposed as an MCP server. Any MCP-compatible client (Claude Desktop, Cursor, custom clients) can connect to your agent and use its tools. You control which tools are exposed and which clients can connect.

### Can I self-host agents on my own infrastructure?

Enterprise plans include a self-hosted option. You get the full platform runtime as a Helm chart that deploys to your Kubernetes cluster. Same API, same features, your infrastructure. Contact sales for details.

### What multi-agent patterns are supported?

The platform supports hierarchical delegation (manager agents assign tasks to specialists), parallel fan-out (one agent triggers many), sequential pipelines (agents pass results forward), and collaborative loops (agents iterate together). All patterns are configurable via the API.

### What is the latency for agent responses?

Median API response time is under 200ms for the platform layer. End-to-end latency depends on the LLM provider and tool calls your agent makes. Streaming responses are supported for real-time applications.

### Can I choose which LLM model my agent uses?

Yes. Agents support OpenAI, Anthropic, Google, and open-source models. You configure the model per agent. Switch models without changing any other code. The platform handles provider abstraction and failover.

### Is the platform SOC 2 compliant?

Yes. Sistava is SOC 2 compliance aligned and not formally certified yet. All data is encrypted at rest and in transit. We support SSO, audit logs, and data residency requirements. Enterprise plans include a dedicated security review and BAA if needed.