Agent workforce platform
Software that runs a team of AI agents as declared, versioned files applied to a database, with approvals and an audit trail built in.
Last verified
An agent workforce platform is software that runs a team of AI agents as an operating unit — agents, their skills, and the policies that govern them are declared in files, applied to a database, and executed by a runtime that records what happened — rather than as one-off scripts wired together by hand. It is distinct from an agent framework, which gives you a graph API and leaves hosting, approvals, versioning, and reporting to you.
What it includes
Beyond an agent runtime, the category implies: a place to author agents, skills, and workflows as reviewable files; a way to compose default agents without forking them; a pause for a human before an action takes effect; a record of which version of an agent produced a given output; and a queue where pending work surfaces for review.
How Vocion implements it
Vocion authors these as YAML and Markdown — workspace manifest, agent, team,
skill, playbook, mission, workflow, automation, object type, source, trust
rules, learning step, eval dataset, workspace page — documented one page per
entity in docs/entities/.
The full inventory, with the table where each entity is authored, its schema,
its table, and its runtime mount, lives in
docs/object-model.md.
workspace:check validates a workspace before workspace:apply writes it to
Postgres and records a workspace_version audit row
(docs/workspace.md).