AI Integration Platform
2kw.ai Documentation
Define an agent once (model, instructions, tools, knowledge and approvals) and 2kw.ai runs it: through an OpenAI-compatible API, inside your own product, or in the chat app. Gateway, traces and evals sit under every agent.
Quick Start
Install the CLI and set your key
Create a key on the API keys page, then:
npm install -g @2kw/ai
export AI_2KW_API_KEY=sk_your_api_key
Create an agent
backbone agents create --name first-agent --model gpt-5.1 \
--instructions "You are a concise assistant. Answer in two sentences at most."
Talk to it
backbone agents run first-agent "What can you help me with?"
The full guide adds the API call, conversations and the chat app: Quickstart: your first agent.
Managed agents
Agents inside your product
Drop a managed agent into your own app. Your UI and brand. 2kw.ai runs the rest.
Your team, your agents
Talk to your organization's published agents in the chat app, with attachments, citations and tool approvals.
Resources
Agents
Versioned assistants with server-side tools, knowledge search, citations and human approval, run through the Responses API.
Agent Tools
Built-in tools, your own webhooks, client-executed tools, knowledge search with citations and files in a turn.
Agent Connectors
Give an agent the tools of an MCP server, with sign-in, approved hosts and images and files in results.
Agent Approvals
Decide which tool calls run on their own and which wait for a person, and set the conversation mode.
Microsoft 365
Let agents read and write Outlook mail and calendar and read contacts, each member signed in with their own Microsoft account.
Chat App
Talk to your published agents at chat.2kw.ai, with attachments, citations and tool approvals.
Scheduled tasks
Have an agent run on a schedule, such as a daily digest, and find each result in a new chat.
AI Gateway
One API, any provider. Use the OpenAI SDK you already know — 2kw.ai routes to the right backend.
Providers & BYOK
Connect your own OpenAI, Azure OpenAI, Anthropic, xAI, Mistral or Ollama account and call it through the gateway.
Data Extraction
Define a JSON schema, feed in any text or document, and get back validated structured output.
Extraction Schemas
Create, version, label, validate and test the JSON Schemas that drive an extraction.
n8n workflows as tools
Let an agent run the workflows of your own n8n, and choose per agent which ones.
Usage & Rate Card
What 1x, 5x and 10x cover, worked examples per model, and the price of every token, page and minute.
Queues
Route traces, spans and experiment results to human reviewers and record their scores against your own rubrics.
Experiments
Compare models, prompts, schemas and parameters against the same test data before you ship.