TengoChatDevelopers
New: agents that use GitHub, Notion, Slack and any MCP server

Build with the AI behind TengoChat.

One API for chat, agents that use your tools, and files served from tengochat.com. It speaks the OpenAI format, so the code you have already works. Pay only for what you use.

  • 2,000 free credits to start
  • No card to try
  • Your prompts are not kept for training unless you opt in
Response
Works with the tools you use
  • OpenAI SDKs
  • LangChain
  • Vercel AI SDK
  • LlamaIndex
  • cURL and any HTTP client

Three ways in. Pick one, or all three.

Start with a single API call, grow into agents that do real work, then put them in front of your users.

Call the API

Chat completions in the OpenAI format, streaming included. Change the base URL in the library you already use.

Use the OpenAI libraries

Build agents

Instructions, a model, tools and connectors, built in Create and tested in a playground. Then call them from anywhere.

Build an agent

Embed anywhere

One script tag puts your agent on any website as a chat bubble, limited to your domains, with a daily spending cap.

Add it to your site

Agents that open your repos, read your docs and post to your channels.

Write what the agent is for. Switch on the tools it may use. Attach the files it should know. Every step shows in the playground, with what it cost.

  • Memory per conversation, kept apart per end user
  • Changes in GitHub or Notion only when you allow them
  • Its own page to share, and a widget for your site
Build your first agent

Your tools, in its hands.

Connect GitHub, Notion, Slack, Discord and webhooks with a token or one click, or add any remote MCP server and its tools appear on their own. Tokens are encrypted and never shown again.

tools from mcp.yourcompany.dev
search_tickets, create_invoice, list_customers
How connectors work

Try it now. No key needed.

A real answer from the API, streamed the way your app gets it.

Ready

The answer appears here, word by word.

The same call with cURL

Files on tengochat.com.

Upload with one call. Share a public address that never changes, or a link that expires. Attach files to agents and they answer from them.

  • 250 MB free, then plans up to 2 TB
  • Downloads included up to ten times your plan each month
  • Byte ranges for video and audio
Store files

Ready for production on day one.

Streaming

Server-sent events for completions and agents: text, tool calls, done.

Threads and memory

Each conversation remembers itself. Send a user id to keep people apart.

Limits you set

Requests a minute per key, monthly spending caps per key, daily caps per public agent.

Usage you can see

Every call with its tokens, credits and time, day by day, by model and agent.

OpenAPI file

Generate a client in any language from the spec, or import it into your tools.

Private by default

Prompts are not kept for training. Only if you opt in, and you can switch it off.

Pay for what you use.

Credits work for everything: models, tools, storage. 1,000 credits are $1, and they do not expire.

ModelInputOutput
tengo-1Balanced: quick and smart, for most apps and agents.$1.00$4.00
tengo-1-miniFastest and cheapest: sorting, short answers, high volume.$0.25$1.00

Per million tokens. Tool calls: 2 credits. Web search: 10 credits.

What would it cost?

About$0a month

Assumes 600 tokens in and 250 out per message.

Full pricing

Help shape our own model. Only if you want to.

We are training TengoChat's own AI. Developers can opt in to let what their agents and API calls say help it learn: names, emails, numbers and keys are taken out first, and switching it off deletes what was waiting. Off unless you turn it on.

How it works

Stuck? Ask the docs.

The docs assistant answers from our guides and reference, with code. Or write to the team.

Ship something tonight.

Make a key, paste one line, and your app has TengoChat's AI.