Your choice of AI tools and agents.
Where your users work today, plus your agents of tomorrow.
Build it in one place. Use it in all of them.
The context layer is the one place you build and maintain what your AI tools and agents know about your CRM data. Deployment is how that reaches everywhere else.
So it has to plug into all of it: the tools your people already work in, the agents you run today, the agents you have not built yet — and the AI work going on alongside them, like an Agentforce agent someone is assembling or a prompt template that needs the same definition of at risk.
Connecting one is short admin work: register it, scope it to a domain, and it starts serving that domain's entities, briefs and skills. It receives that domain and nothing more, and never more than the person asking could already see.
It reaches past your org too. Definitions export as a portable bundle and as an Apache Ossie semantic model — the vendor-neutral standard Salesforce itself contributes to — so what you build here can be read by other platforms.
What happens on a request
Agentforce
Claude
ChatGPT
Slack
Teams
Any AI agent Connects through Salesforce's own mechanisms.
Native
Hosted MCP servers, invocable actions, prompt grounding — every route in is a capability Salesforce has.
Secure
Salesforce's own OAuth and registry, plus record and field security on every call.
Open
Any MCP client, any agent platform, and whatever you adopt next.




