Skip to main content
RaisePill is autonomous, but you are always in charge. Every mechanism below exists to make sure the agent never does something you didn’t sanction.

​
Autonomy levels

Each project has an autonomy setting that controls how much the agent can do without asking:
LevelWhat the agent can do without you
ConservativeRequires approval for every external action
BalancedCan draft and research autonomously; asks before sending anything
AggressiveCan send investor outreach within your defined budget cap
Change the autonomy level in Project Settings → Autonomy.
Regardless of autonomy level, grant submissions and warm-investor replies always require your explicit approval. This cannot be changed.

​
Hard constraints

Hard constraints are rules the agent will never break, regardless of autonomy level. You define them in Project Settings → Hard Constraints. Examples: 
Never email Tier-1 funds without my approval
Never submit applications without my approval
Skip all consultant outreach
Monthly outreach budget cap: 50 emails
When the agent encounters a situation covered by a hard constraint, it creates a Task instead of acting.

​
Sources and training

The agent only works from sources you have explicitly attached to your project. It never:
  • Pulls data from other projects or workspaces
  • Uses data from outside your Data Room
  • Trains on your data for general model improvement
You can review every source attached to your project and revoke any source at any time from Project Settings → Data Room.

​
Audit log

Every agent action is logged with:
  • Timestamp
  • Which agent acted
  • What action was taken
  • The full reasoning chain (what data was used, what decision was made)
The audit log is visible on the Agent page activity feed and can be exported as CSV or JSON for compliance purposes. Log retention is configured at workspace level in Organization Settings → Audit log retention.
Last modified on May 18, 2026