An agent that already worked on the project receives only what changed since then. When it calls open_project, SyncNow's server finds the previous work session of the same team member on that project, from any PC, and the moment it ended (or its last operation): the answer is then a delta with only what changed since that moment in Planning (plans added, changed or completed), the Rules (rules added, changed or removed), Activity (relevant entries, with who, what and how many files) and Conversations (new questions and answers), each with counts and a line saying where to read the whole section.
It receives the full context instead, capped as always, the first time the person works on the project, when the previous session ended more than 30 days ago, when that session has no readable time, when the delta cannot be built, or when an agent on SyncNow 13.2.49.63 or later asks for it with context set to full; the answer always says which mode it is, from when and why. Whatever the mode, the agent always receives all the rules of the environment it said it works on, or of both Staging and Production when it did not say, whatever files it touches, the blocking questions still open, the other live sessions on the same files, the person's unfinished work and the questions waiting for them. The server decides for every version of the agent tools.
To read more, agents use filters like the app's. get_planning and search_planning filter by state, category, environment, priority, person, period and words; get_environment, which reads the Rules, by section (Staging or Production), person, period and words, refusing filters that rules do not have, such as category or priority; syndication_list by state, priority, who asked, period and words; and project_activity reads the Activity with the app's Filters (type, period such as 7d or 30d, file or path, area) plus person and words, page by page. Every answer is capped and says when it was cut. See What does my agent know about the project before it changes anything?.