nousc/opennous

Opennous, also called Nous, is an open source project that gives AI agents a shared memory about your sales and marketing accounts. Instead of an agent having to check your CRM, inbox, calendar, LinkedIn, and other tools separately every time it needs to know about a customer, Nous pulls all of that scattered information into one place and hands it back in a single request. The project calls this a context graph, and it works in three layers. First, every piece of information from your connected tools becomes an observation, a permanent record of something that happened or was said. Second, those observations get matched to a single canonical record for each person or company, called an entity, so duplicate contacts and profiles are merged into one. Third, from all the evidence attached to an entity, Nous derives claims, which are the current best beliefs about a fact, each one marked with how confident the system is and how fresh the information is. Nous exposes a small set of endpoints an agent can call. get_context returns the whole picture for one account or person, sized to fit within a token budget. get_account returns the full record for a single person or company. query lets you search and filter activity across many accounts at once, optionally ranked by a plain language question. Getting started is a single command, npx @opennous/cli@latest init, which works the same way on macOS, Linux, and Windows as long as Node.js 18 or newer is installed. It signs you in, connects Nous to your coding agent, and can also point to a self hosted instance if you run your own server. The project targets teams building AI agents for sales and go to market work who want their agents to act on consistent, sourced, up to date account information rather than piecing it together from six different tools each time. It is released under the AGPL v3 license, and a hosted version is also available.