Product milestones and platform updates. For the deeper story behind each, see the blog.
Build, run, and govern are now unified in one platform. The orchestrator builds your agent systems; you run and govern them live by talking to it in plain language. Every agent runs as a self-identity, every action lands in the immutable audit log, and any action can be replayed or issued as an audit packet on demand.
Agents that author agents went live in production on 3 June 2026. The lead orchestrator captures intent, and the product specialists write the requirements, build the agent and its workflows, and instrument it for observability. The authoring work itself is now automated.
Conversational, workflow, and research runtimes are live and shipping in production — the foundation for Concierge, Operator, and Analyst agents.
Persistent, judge-gated agent memory went live through the Context Service — append-only, scoped global → agent → session — carried at runtime as CS-packets.
Every agent action writes to an immutable, append-only audit log with a signed, PII-free receipt; PII is tokenized before any model call; and any action can be replayed or issued as an audit packet on demand.
The Agent OS is live. Ten agents now run in production — five build and run customer agent systems, five run Olbrain itself.
SOC 2 and ISO 27001 programs in progress; India data residency (Google Cloud, Mumbai) and DPDP-aligned architecture. Current control status is kept honestly labelled on the Trust page.