# AgentBarrier, Release Stories AgentBarrier is a deterministic test harness for the control guarantees around AI-agent tool execution. It verifies that approval, rejection, cancellation, timeout, replay, delegation, ambiguous outcomes, audit receipts, and parallel execution controls prevent unintended side effects. ## AgentBarrier 1.0.0: keep a human in the loop when AI agents want to take action _v1.0.0 · 2026-08-21_ AgentBarrier now lets you approve or block what your AI agents do, with dashboards, Slack, and secure record-keeping built in for production use. ### What changed AI agents that can call tools or take actions on their own are powerful, but that power needs a check. AgentBarrier 1.0.0 introduces a core enforcement system that decides, based on rules you set, whether an agent's action should proceed, and it keeps a record of that decision. This works whether your agents are built with OpenAI Agents, LangGraph, PydanticAI, or Google's Agent Development Kit, since each of these frameworks now has its own connector to AgentBarrier's approval system. There's also a gateway for agents that use the MCP protocol, so those tool calls are checked and validated too. To act on these approvals, you get a secure web dashboard where pending requests show up in real time, complete with an audit trail of what happened and when. If your team prefers to work from Slack, you can review and approve or deny requests right from there instead of switching tools. Approval events can also trigger signed webhooks, so other systems can be notified reliably and the notifications can be verified as authentic. For teams running this in production, AgentBarrier now supports PostgreSQL as a storage option alongside SQLite, so audit logs and approval records can scale with your setup. Access to the dashboard and approval workflows is controlled by role-based permissions, so different team members can have different levels of authority. Operators also get controls to pause, resume, or limit how much an agent runtime can do, giving a way to step in if something needs to stop. This release also fixes a privilege configuration issue in the PostgreSQL storage backend and strengthens how the system recovers from errors and validates policy rules, so the enforcement behavior stays consistent and predictable. ### Why it matters With this release, approving or blocking an AI agent's actions is no longer something you have to build yourself. You can see what agents are asking to do, decide from a dashboard or Slack, and trust that the record of those decisions is secure and auditable, whether you're running one agent framework or several. ### How to use it Connect the adapter for your agent framework (OpenAI Agents, LangGraph, PydanticAI, or Google ADK) so its tool calls route through AgentBarrier. From there, review and act on pending approvals using the web dashboard or directly in Slack, and set up webhooks if you want other systems notified when approval decisions are made. ### New Features - Add runtime approval enforcement core, Introduces the foundational approval enforcement mechanism for the AgentBarrier runtime, enabling policy-driven control over agent tool execution with deterministic rule evaluation and state tracking. - Add production MCP policy gateway, Implements a production-ready MCP protocol gateway that enforces policies on tool calls from MCP-compatible agents, hardening the transport boundary and validating policy compliance at the gateway level. - Add authenticated approval service, Provides a secure, authenticated service for managing tool approval workflows, including multi-user authorization controls and signed approval contracts. - Add durable signed webhooks, Enables webhook-based notifications for approval events with cryptographic signing to ensure integrity and auditability of approval state changes. - Add OpenAI Agents runtime tools, Adds adapter for OpenAI Agents framework, enabling AgentBarrier policy enforcement on tool calls from OpenAI-based agents. - Add LangGraph runtime tools, Adds adapter for LangGraph framework, enabling AgentBarrier policy enforcement on tool calls from LangGraph-based agents. - Add PydanticAI runtime tools, Adds adapter for PydanticAI framework, enabling AgentBarrier policy enforcement on tool calls from PydanticAI-based agents. - Add Google ADK runtime tools, Adds adapter for Google Agent Development Kit, enabling AgentBarrier policy enforcement on tool calls from Google ADK agents. - Add secure approval dashboard, Provides a web-based dashboard for managing and approving pending tool calls with real-time status visibility and audit trail. - Add PostgreSQL runtime storage, Adds PostgreSQL as a storage backend option for audit logs and approval state, complementing SQLite for production deployments. - Add secure Slack approval workflows, Integrates Slack for approval notifications and workflow control, allowing approvers to review and act on pending tool calls directly from Slack. - Add multi-user authorization controls, Implements role-based access controls and multi-user permission management for the approval dashboard and API. - Add production runtime safety controls, Introduces operational controls including pause, resume, and execution limits for managing runtime behavior in production environments. ### Bug Fixes - Harden runtime recovery and policy validation, Strengthens error recovery mechanisms and adds comprehensive policy schema validation to ensure deterministic rule enforcement. - Harden MCP gateway transport boundary, Reinforces security at the MCP gateway level with additional validation and isolation of untrusted agent inputs. - Fix PostgreSQL privilege audit, Corrects privilege configuration in PostgreSQL storage backend to ensure proper access control enforcement. ### Maintenance - Add strict runtime policy schema, Defines and enforces a strict schema for policy-as-code YAML rules to enable early validation and prevent configuration errors. - Add runtime database operations, Implements core database operations for managing approval state and audit records in the SQLite-backed runtime. - Finalize runtime approval contract, Establishes the final approval API contract and integration points for runtime enforcement. - Refresh development compatibility evidence, Updates compatibility documentation and test matrices for supported agent frameworks and Python versions. - Audit runtime lifecycle from built wheel, Validates runtime behavior and security properties through comprehensive testing of packaged distributions. - Complete production control plane safeguards, Finalizes all security and reliability features for production deployment of the approval control plane. - Enforce PostgreSQL coverage gate, Establishes test coverage requirements for PostgreSQL storage backend to maintain reliability standards. - Complete 0.3.0 roadmap, Fulfills all planned features and improvements for the 0.3.0 release cycle. - Prepare AgentBarrier 1.0.0 release candidate, Finalizes all breaking changes, features, and documentation for the 1.0.0 stable release. - Use CLI-first verified PyPI releases, Adopts CLI-based verification for PyPI package integrity before release publishing. - Make release publishing idempotent, Improves release publishing process to ensure safe, repeatable operations without side effects. ## AgentBarrier 0.3.0: More Flexible Approvals and Better CrewAI Support _v0.3.0 · 2026-08-21_ This release makes it easier to set up custom approval workflows, adds a payment tracking example, and fixes CrewAI compatibility on Python 3.10. ### What changed AgentBarrier now supports approval barrier profiles, which let you create different approval workflows for different scenarios without changing code. This gives teams more flexibility in how they configure agent approval processes to match their specific needs. The update also improves how the system handles unclear or ambiguous decision outcomes, making agent decision validation more robust and predictable when situations aren't clear-cut. Alongside this, a new example has been added showing how to use agent barriers with a SQLite-backed payment ledger, giving developers a working reference for payment tracking use cases. For CrewAI users, this release adds deterministic evaluation, meaning agents can now run repeatable tests that produce the same results every time. This release also restores CrewAI support for Python 3.10 environments, fixing a compatibility issue that had prevented it from working correctly. Finally, compatibility evidence is now automatically generated and validated during continuous integration, helping ensure consistent support across versions before changes reach users. ### Why it matters Together, these changes make AgentBarrier more adaptable and dependable. Teams can configure approval processes to fit their own workflows, trust that ambiguous decisions are handled more reliably, and rely on consistent, repeatable behavior when testing with CrewAI. The Python 3.10 fix also means CrewAI users on that version can resume using AgentBarrier without workarounds. ### How to use it To take advantage of the new flexibility, set up approval barrier profiles to define different approval workflows for different scenarios. If you're working with payment tracking, refer to the new SQLite payment ledger example for a complete working reference. CrewAI users can now run deterministic evaluations to get consistent, repeatable results across test runs. ### New Features - Add approval barrier profiles, Introduce approval barrier profiles to enable more flexible and configurable approval workflows within agent barriers. - Generate compatibility evidence in CI, Automatically generate and validate compatibility evidence during continuous integration to ensure consistent cross-version support. - Expand ambiguous outcome reconciliation, Enhance reconciliation logic to handle a wider range of ambiguous outcomes, improving robustness of agent decision validation. - Add SQLite payment ledger example, Include a new example demonstrating how to use agent barriers with SQLite-backed payment ledger implementations. - Add deterministic CrewAI evaluation, Implement deterministic evaluation capabilities for CrewAI integrations to ensure reproducible agent behavior. ### Bug Fixes - Restore CrewAI support on Python 3.10, Fix compatibility issues to re-enable CrewAI support for Python 3.10 environments. ## AgentBarrier now shows the correct version number _v0.2.1 · 2026-08-21_ A small fix ensures AgentBarrier accurately reports which version you're actually running. ### What changed Previously, AgentBarrier could report an incorrect version number in published releases, which meant the version displayed didn't always match what was actually running. This release corrects that issue. AgentBarrier now accurately reports its published version. ### Why it matters With this fix, your application now correctly displays the actual published version, so you can trust that the version information you see is accurate. ### Bug Fixes - Fix published version reporting, Corrects an issue where the published version was not being reported accurately.