Developers ship one endpoint. Your team tunes the agent live. No deploy, no ticket, no waiting on engineering.
Bring your own model. Works with the providers you already use.
Stop shipping a release every time the agent needs a tweak. LiteAgents splits the work so neither side waits on the other.
Start a conversation, send a message, get the reply. The same call whether one agent answers or five work together behind it.
# start a conversation with an agent… POST /agents/{agent}/conversations # …then send and list its messages. POST /conversations/{conversation}/messages GET /conversations/{conversation}/messages
Fix a bad answer, tighten a guardrail, swap a tool. It goes live the moment you save. Versioned, testable, reversible.
Instructions, guardrails, tools, model. All editable, all versioned. The agent that just gave a bad answer is one edit away from a good one. No release cycle.
Upload your docs, and it answers from those. No guessing. Files and wiki pages live in a Library any agent can read. Production RAG, no vector database and no embeddings to tune.
Any HTTP request becomes a tool with
{{context.*}} variables. Connect MCP
servers over OAuth and their tools show up on their own. Flip on built-in connectors. Secrets stay encrypted,
never in your repo.
No new engine to learn. A sub-agent is just another tool a parent can call. Wire a parallel team or a sequential pipeline, all in the same panel. Your one endpoint never changes.
Compose a workflow
The logic and tools live in LiteAgents, not your codebase. Your repo stays clean while the agent keeps getting smarter. Zero deploys on your side.
// 1. open a conversation for your end-user const { conversation } = await post( `/agents/${agent}/conversations`, { user_id } ) // 2. send whatever they typed const reply = await post( `/conversations/${conversation}/messages`, { text } ) // 3. render reply.text + reply.actions // the brain lives in LiteAgents.
We serve as your bridge. Pick your own preferred AI provider and bring your own provider keys for your agents. Pay us for the platform, pay your provider for tokens. Nothing in between.
Every plan includes unlimited agents. Estimate your run volume, bring your own model keys, and never pay a markup on intelligence.
*Subject to Fair Usage Policy of the Terms of Use.
Integrate once. Tune forever. Free to start, bring your own key.
