Party's per-agent model choice and session boundaries distinguish its shared build workflow from OpenAI's cloud-run, Codex-powered organization-shared agents: the team selects a harness and model for each agent, and Party's environments define the runtime and boundaries of each session, with tenant-level sandboxing adding a further layer of isolation around the team's work.
OpenAI designed its workspace agents so that they can be used in either ChatGPT or Slack.
Party's October team-building activity could help it attract interest in multi-step agent workflows, while workspace agents OpenAI has made generally available raise the competitive bar; repeat team use after Party events would confirm whether such activity converts into growth. Frontier Professionals use agents for multi-step workflows and for building multi-agent systems, workspace agents are now generally available across ChatGPT Business, Enterprise, and Edu, and teams put Party to work under deadline pressure on October 10.
At Party's October 7 build event, agents handled planning, coding, and review, while people kept control of direction and decisions.
Trust is a material adoption risk for Party's team-agent workflow: in PwC's May 2025 survey, 28% of respondents ranked lack of trust in AI agents as a top-three challenge, while 31% trusted agents for daily collaboration with human team members.
Party specifies operational safeguards for its collaborative workspace: every agent session runs in an isolated sandbox with its own filesystem and process space, and workspace roles govern who can start sessions, approve merges, and change budgets.
Party gives teams a shared product-building workspace where agents plan and build in milestones with a live Preview for feedback as work develops, and team conversations from Slack and Teams stay connected to the work.
OpenAI also offers shared agents in ChatGPT or Slack for team workflows; its workspace agents are shared within an organization and can be used in ChatGPT or Slack.