Standards
We support and build on financial industry standards. Boagent is a SWIFT Partner and we encourage our team to spend work time participating in Open Source development and contributing to better industry standards.
Committed to governance, standards, and privacy
Go to Trust Center 
Governance
Boagent natively audits, monitors, and attributes every action to users, agents, and agents-on-behalf-of users for SEC and FINRA compliance.
Privacy
Gemma
GLM
MiniMaxWe are model agnostic and only work with inference providers and models that allow for minimum privilege and zero retention data policies. Boagent supports Open Source models and offers managed services for private deployments via Modal AI.
These principles are engrained in Boagent’s code
Before the work
Make the process explicit.
Give the agent relevant context and approved procedures. Review the flow before publication, and decide which actions can proceed and which require a person.
Policy & Context EngineDuring the work
Keep decisions in reach.
Review the proposed action and supporting evidence. Approve the specific request or reject it with a reason, so the next step follows your decision.
Approval GatewayAfter the work
Keep a clear record.
Follow the decision through to the action’s recorded result. Keep open questions, evidence and proposed treatments together as the investigation progresses.
Investigation & ResolutionA few practical questions
Understand the boundaries.
Who decides what the agent can do?
People manage action settings at the organization and workflow level. Relevant facts give the agent context, and human-approved procedures guide its approach. Neither gives the agent permission to change its own authority.
What happens when an approval is needed?
The agent presents the specific proposed action for review. A person can approve it or reject it with a reason. Approval applies to that action; a changed proposal needs to be evaluated again. A rejected action does not proceed.
Does approval mean the action succeeded?
No. The decision and the execution result are separate records. After approval, the agent can attempt the action and inspect its result. A recorded send result, for example, is distinct from a human approving the message.
What is reviewed before a flow or report goes live?
The agent prepares the draft and runs the available checks. A person reviews publication. Flow checks inspect the process structure; the team still needs to validate its business rules and results against the intended work.
Can we see what happened?
Conversations, workflow runs, action records and approval decisions keep the work traceable. Review the proposed action, who decided, and the outcome reported by the tool. Discuss retention and any specific audit requirements with our team.
How does setup work?
We work with your team to define the job, scope the available data and communication channels, review the process and set action policies before putting it into operation.
Where can we discuss security requirements?
Contact our team for a review of deployment, access, data handling and any requirements specific to your organization. We will work through the relevant details with your security and operations teams.
Bring us your security, governance and operational requirements. We’ll work through the details together.
Talk to our team