Build, Scale, and Monitor
with serverless cloud hosting, vector search, webhooks, callbacks, and more.
Tools with Memory
Build serverless agents with our low-code Python SDK.
- Create personalities with prompts
- Generate images with multi-modal support
- Reason & Use Tools with ReAct
- Connect to apps with API endpoints
Robust LangChain support built-in.
class StarTrekCaptainAgent(TelegramAgentService): def __init__(self, **kwargs): super().__init__(incoming_message_agent=None, **kwargs) self.incoming_message_agent = ReACTAgent( tools=[ StableDiffusionTool(), VectorSearchQATool(), ], llm=OpenAI(self.client), ) self.incoming_message_agent.PROMPT = self.build_dynamic_prompt()
Run pre-made agents in the cloud
When you deploy an agent, anyone can create a private instance to use on the web, Telegram, or as an API.
- Describe a personality to create a GPT bot that lives it.
- Generate DALL-E images by saying 'dalle <prompt>'.
- Upload PDFs and YouTube videos on the web. Then ask questions about them in Telegram.
- Generate spoken audio by saying 'say <prompt>'.
Create a Character
Voice Gen Bot
Scale fleets of agents with ease
- Growth: Can I support 1M Users?
Yes. Every agent auto-scales from 1 to N.
- Optionality: Can I swap different LLMs?
Yes. LLMs, Image Generators, Transcription engines.
- Vector Search: Do you support it?
Yes. With our integrated Vector Engine & 3rd party options.
- Multi-User Tehancy: How do I offer it?
Every Steamship Agent is multi-tenant by default.
- Versioning: Is this new code different?
Know by keeping track of prior versions.
Monitor your agents and their behavior
- Task History: What is my agent doing?
Review every LLM call, reasoning step, and image generation.
- Logging: Why did it behave oddly?
Know with detailed invocation logs.
- Usage: How much is it costing?
Consolidated tracking across AI providers.
Consolidate training, inference, vector search, and endpoint hosting.
Build and deploy AI projects with confidence.
$10/ mo USD
+ Model Costs
- Multi-modal agents
- Managed vector database
- Async task chains
- Persistent user state
- Cloud agent logging
- Python API endpoints
- OpenAI, Cohere, HuggingFace, Eleven Labs, AssemblyAI, and more
- Third party models billed as-used & at-cost
- Key management, load balancing, and traffic shaping is automatic
- Detailed per-model & per-agent usage reporting
- Discord community support
- Agent building guidebook
- Growing open-source template library