Maya agent runs, under the hood: plan, tools, memory, approve
Ask Maya for a week of posts and you get back a finished campaign pack — but what happens in between is a deliberate, auditable loop. Every run starts with a stored plan, each step calls one logged tool, runs build on your business profile, and the run always stops at awaiting-approval.
Plan first, tools second
Maya decomposes your goal into an ordered plan stored with the run before any tool executes. Each step then calls one real MarketFlow capability — campaign strategy, post generation, scheduling drafts, keyword research — and every call and outcome is recorded as a step on the run.
Memory makes runs compound
Runs read your business profile, past approved campaigns and voice settings before drafting, which is why captions reference your services and city instead of generic filler. Memory is scoped to your workspace and never shared across customers.
The run always stops for you
Completed runs halt in an awaiting-approval state with the full pack visible: strategy, captions, visuals, schedule. There is no code path from a completed run to a published post that skips this gate — publishing is a separate, human-triggered action.
Frequently asked questions
- Can a Maya agent run publish on its own?
- No. A run ends at awaiting-approval with a draft pack. Publishing is a separate action you trigger after approving, and only to accounts you connected yourself.
- What tools can Maya call inside a run?
- Only capabilities that already exist in your Studio: campaign strategy, post and caption generation, image drafts, scheduling drafts, keyword research and listing checks. Each call is logged as a step.
- Where can I see everything a run did?
- In your Studio. Each run shows its plan, every tool step in order, and the final pack awaiting approval.