Managed policies, plus your own.
Octic-managed policies with enhanced protections apply out of the box, and your custom policies run alongside them. Together they evaluate every AI agent, MCP server, and coding agent, set each one’s trust state, and turn a violation into a finding with an owner.
Policies evaluating Acme Corp’s MCP servers
Every MCP server has an owner.
mcp-servers-need-ownerCustom, Acme Corp
MCP servers with file or database access need a person’s review.
sensor/mcp-data-accessOctic-managed, sensor pack
1 of 3 MCP servers is flagged and becomes a finding with an owner.
mcp-jira
Found via API
Owner @platform
mcp-servers-need-owner: Approve
sensor/mcp-data-access: Pass
Approved
mcp-filesystem
Found in network traffic
Owner @platform
mcp-servers-need-owner: Approve
sensor/mcp-data-access: Review
Unreviewed
mcp-postgres-prod
Found in network traffic
Owner No owner
mcp-servers-need-owner: Flag
sensor/mcp-data-access: Review
Flagged
- Rule
- mcp-servers-need-owner
- Reason
- No owner derived
- Finding
- MCP server with no owner
- Owned by
- Lena Okafor
- SLA
- 14 days
- History
- Flagged, then assigned
On Acme Corp's map, each MCP server takes the trust state its verdicts add up to: mcp-jira approved, mcp-filesystem unreviewed, mcp-postgres-prod flagged. mcp-postgres-prod has a finding, mcp server with no owner, owned by Lena Okafor (@platform-leads).
Managed policies apply before you write a rule
Octic maintains managed policies, including enhanced protections, in one pack per source: your cloud, your code host, your model providers, Octic’s sensors, and more. Each pack is versioned, and you can turn off any rule in it.
Custom policies say what your organization allows: each one approves what it matches, or flags it. They run alongside the managed packs, and Octic evaluates every AI agent, MCP server, and coding agent against both.
- Octic-managed rule sensor/corporate-accounts, in the sensor pack: Coding agents sign in with a corporate account.
- Octic-managed rule sensor/mcp-data-access, in the sensor pack: MCP servers with file or database access need a person’s review.
- Codex fails sensor/corporate-accounts: it is signed in with a personal account. Flagged.
- mcp-filesystem waits for a review under sensor/mcp-data-access: it reads local files. Unreviewed.
- Acme Corp's custom policy mcp-servers-need-owner runs alongside them: Every MCP server has an owner. mcp-postgres-prod has no owner, so the policy flags it. Flagged.
The managed rules decide Codex and mcp-filesystem. The custom policy is what makes mcp-postgres-prod flagged.
Start from a template, and see what a rule would flag
Build custom policies in the visual builder, starting from a template. The impact preview shows which AI agents, MCP servers, and coding agents the rule would approve or flag.
Each policy shows an example asset that triggers it and one that complies, as JSON, so you can check the rule does what you meant.
- Started from the template Require an owner: applies to MCP Server; requires an owner that is set; otherwise, flag.
- Example asset that triggers the rule: mcp-postgres-prod, an MCP Server with owner null.
- Example asset that complies: mcp-jira, an MCP Server with owner @platform.
- Impact preview: the rule would flag 1 of 3 MCP servers.
- mcp-jira: owner @platform. Approve.
- mcp-filesystem: owner @platform. Approve.
- mcp-postgres-prod: no owner. Flag.
A violation becomes a finding with an owner
When an AI agent, MCP server, or coding agent fails a policy, Octic sets its trust state and records the violation as a finding.
Every finding has an owner, an SLA, and a history, so each gap a policy finds lands with a person, a deadline, and a record of what happened to it.
- Acme Corp's custom policy mcp-servers-need-owner evaluates mcp-postgres-prod, which shadow-summarizer connects to.
- mcp-postgres-prod has no owner, so the policy flags it. Flagged.
- The violation becomes a finding: mcp server with no owner. Status open.
- The finding is owned by Lena Okafor (@platform-leads).
mcp-postgres-prod is flagged, and its finding is owned by Lena Okafor, who leads Platform.
Where this sits in Trust
The policy engine sets each trust state. The risk register gives you one risk view across your AI estate, including what your policies flag.
Back to TrustStart with the managed packs, then add your own
Sign up and Octic-managed policies apply once your estate is on the map. Build a custom policy from a template, and see which AI agents and MCP servers it flags.