SUBJECT Signal Brief: Agents get their marketing API, a shared memory, and Google Cloud's work agent
Zernio
Zernio — launched October 9 — is marketing infrastructure for products and agents: one integration on official platform APIs covers 16 social platforms, ads on 7 networks, a unified inbox, SMS, RCS, iMessage, WhatsApp, and calls and numbers in 57 countries — and a hosted MCP server and CLI let agents connect accounts, run campaigns, send messages, and place calls on their own — with free options.
WHY IT MATTERS Roadmaps now carry an agents-talk-to-customers line item — one API that absorbs the per-channel integrations, platform approvals, and maintenance turns a quarters-long channel program into a same-day launch, and the hosted MCP server is the distribution pattern agents will expect from every marketing tool.
zernio.com →Busabase
Busabase — launched October 9 — is a general-purpose database and workspace built as the system of record for a team's AI agents: business records, docs, skills, and apps live in one shared base that Claude Code, Codex, and more connect to — with access controls, change review when needed, and a visible history of what every agent changed — open source and free, in cloud, personal-desktop, and self-hosted options.
WHY IT MATTERS Agents that cannot share context redo each other's work — a system of record built for agents, with the access controls and audit trail enterprises demand, is the memory layer PMs will build or buy before their agent features can compound instead of collide.
busabase.com →Together Link
Together Link — launched October 9 — connects the coding agents a team already runs — Claude Code, Claude Desktop, Codex, ChatGPT Desktop, OpenCode, and Pi — to open models on Together AI through a one-command setup: keep the existing agent, login, settings, and history, pick models like Kimi K3, GLM 5.3, and DeepSeek V4.1 Flash or let Auto Router route each task, and get a cost receipt for every session — free and open source, with per-token billing.
WHY IT MATTERS Model choice is becoming a routing decision instead of a migration — a drop-in link that swaps coding agents onto open models without touching the workflow makes cost-per-session a visible, optimizable number, which is exactly the lever PMs owning AI budgets have been missing.
together.ai →Odyssey 3
Odyssey 3 — launched October 9 — is a foundation world model that generates interactive environments from a prompt and predicts in real time how they change as a person or an agent acts — its Pro version posts the highest reported Physics-IQ Verified video-to-video score (66.1), and the same model has been adapted to control robot arms and humanoids, drive a car, and train agents — free research preview, with API access on request.
WHY IT MATTERS Testing against the physical world has meant staged environments and special-effects budgets — a prompt-generated, physics-accurate simulator turns real-world testing and agent training into an on-demand API call, rewriting the build-vs-buy math for PMs shipping into robotics, operations, or spatial product surfaces.
odyssey.systems →Gemini Agent for Google Cloud
Gemini Agent for Google Cloud — launched October 9 — is Google Cloud's single, universal agent for work: give it an objective instead of instructions and it plans, connects to Workspace, Microsoft 365, Slack, Salesforce, Jira, or any MCP server, and hands back finished docs, decks, or code — running in the cloud for hours or days, spinning up coworker agents with their own email and calendar, and routing each job across Gemini and Claude models under hard spend caps — with free options.
WHY IT MATTERS The frontier platforms are now shipping the agent, not just the model — an enterprise work agent that runs for days, spawns coworker agents, and model-routes under hard spend caps is the reference architecture PMs will be benchmarked against, and the clearest signal yet that agents get governed like employees, not autocomplete.
cloud.google.com →