Bulwark
Run policy checks, content scans, and audited agent sessions when you need guardrails around what Claude or Codex can say and do in production.
Overview
Bulwark is a MCP server for the Ship phase that governs AI agents with content scanning, policy evaluation, audit logs, and session management.
What is this MCP server?
- Content scanning on agent inputs and outputs against configurable policies
- Audit logs for operator review and compliance trails
- Policy evaluation engine driven by bulwark.yaml configuration
- Session management with operator tokens (bwk_sess_...) for authenticated control
- stdio MCP transport via OCI image ghcr.io/bpolania/bulwark:v0.2.0
- Server version 0.2.0
- stdio transport via OCI image ghcr.io/bpolania/bulwark:v0.2.0
- Two configured environment variables: BULWARK_CONFIG and BULWARK_SESSION_TOKEN
Community signal: 4 GitHub stars.
What problem does it solve?
Autonomous coding agents can leak policy-breaking content or run without a traceable session boundary unless you bolt on governance outside the chat UI.
Who is it for?
Indie builders shipping customer-facing agents who need lightweight policy and audit hooks without building a custom compliance platform.
Skip if: Teams that only need local linting on code or who have no agent surface area requiring content or session governance.
What do I get? / Deliverables
After you register Bulwark, your agent stack can enforce policies from YAML, log decisions, and tie operator actions to authenticated sessions.
- Policy-evaluated agent interactions with configurable scanning
- Persisted audit log stream for governance review
- Managed operator sessions bound to secret session tokens
Recommended MCP Servers
Journey fit
How it compares
Agent governance MCP server with audit and policy tooling, not a general-purpose LLM proxy or a static security scanner skill.
Common Questions / FAQ
Who is Bulwark for?
Bulwark is for solo builders and small teams running Claude Code or similar agents who need scanning, policies, and audit trails around autonomous sessions.
When should I use Bulwark?
Use Bulwark when you are in Ship or hardening agent workflows and need configurable rules, logged evaluations, and operator session tokens before going live.
How do I add Bulwark to my agent?
Add the stdio MCP server using the OCI package ghcr.io/bpolania/bulwark:v0.2.0 with arguments mcp start, set BULWARK_CONFIG to your bulwark.yaml path, and supply BULWARK_SESSION_TOKEN for operator auth.