Build on the engine.
Every capability is an API. Run auto-AGENTS™ from your own stack with streaming inference, MCP-native tool use and 170+ connectors, no required UI. Your agents call ours; ours call your tools. Model-agnostic, low-latency, and deployable in cloud, VPC or on-prem.
Every capability is an endpoint.
There is no UI you must adopt. Agents, runs, tools, knowledge, voice, evals and analytics are all exposed as a clean REST + streaming API. Drive auto-AGENTS™ from your own application, your IVR, or your own agents, the engine is headless by design.
Your agents call ours. Ours call your tools.
auto-AGENTS™ speaks the Model Context Protocol natively in both directions. Expose your own tools and the engine will reason over them; register ours and your agents gain payments, CRM, ticketing and data access, every call grounded, governed and replayable in a full trace.
Tokens, events and audio, as they happen.
Stream model output token-by-token over SSE or WebSocket and emit structured events for every reasoning step and tool call. Sub-500ms first token keeps voice natural; webhooks deliver run completions, scores and escalations to your systems the instant they occur.
Run it in your cloud, your VPC, or on-prem.
Model-agnostic and portable, bring GPT, Gemini, Claude or your own on-prem Llama. Pin to a region for data residency, isolate in your VPC, or deploy fully on-prem. Enterprise auth via SSO / SAML / OIDC, scoped API keys, and an immutable audit trail on every call.
An engine your builders can stand on.
Connectors, day one
Reach across CRM, payments, ticketing, data and telephony over MCP without writing a single integration adapter.
First-token latency
Streaming inference keeps voice and chat natural, with structured events for every reasoning and tool step.
Concurrent sessions
A horizontally-scaled runtime with admission control, proven at high-concurrency voice volumes.
Certified for regulated builders.
Deterministic guardrails before and after every model call, PII vaulting, scoped credentials and your choice of region, with an immutable audit trail behind every API request.
Questions from developers.
Every capability of auto-AGENTS™, running agents, registering tools, indexing knowledge, voice, evals and analytics, is exposed as a REST + streaming API with no required UI. You can run the entire engine from your own application, your IVR, or your own agents. The dashboard is just one optional client of the same API you get.
auto-AGENTS™ speaks the Model Context Protocol natively. You can expose your own tools as MCP servers and the engine will reason over and call them within your guardrails; you can also register our 170+ connectors so your agents gain payments, CRM, ticketing, data and telephony access. Every call appears in a full run trace, reasoning, tool call, result and timing.
Stream model output token-by-token over SSE or WebSocket, plus structured events for every reasoning step and tool call. First-token latency runs under 500ms to keep voice natural. For asynchronous work, webhooks deliver run completions, QA scores and escalations to your endpoints the instant they happen.
Yes. Bring GPT, Gemini, Claude or your own on-prem Llama, swappable per run with fallbacks. Deploy in our cloud, your private VPC, or fully on-prem, and pin to a region for data residency. SDKs ship for TypeScript, Python and Go.
Enterprise auth via SSO / SAML / OIDC, scoped and rotating API keys, field-level encryption and tenant-scoped key management. Deterministic guardrails run before and after every model call, PII is vaulted, and every API request is recorded in an immutable audit trail. The platform is certified to SOC 2, HIPAA, ISO 27001, PCI and GDPR.
Built directly on the headless engine, agents, tools and knowledge driven entirely over REST + streaming, with every reasoning step and tool call replayable from the run trace.
170+ tools your agents can call on day one.
Build agents on the headless engine.
Every capability is an API. Bring your stack and your models, we'll wire the engine in.
