Changelog
1.8.0
This version brings a big set of features and improvements which will notably improve the experience.
New
Support for multiple providers and changing models dynamically
New UI for agent interaction and chats
Light mode for the interface.
Google Vertex AI support.
Progress visibility for graph processing jobs.
Focused graph views highlighting findings relevant to the engagement scope.
Improved
Easier workspace selection and switching.
Graph clustering and performance for smoother exploration.
More stable live graph updates.
More selective collection of security-relevant information.
Better support for long-running AI conversations.
Clearer representation of hosts and open services.
Simplified terminal menus and fewer unnecessary notices.
Fixed
Workflows and graph views now follow the selected workspace.
Corrected local-time display.
Corrected AI model names shown in responses.
Improved handling of interrupted terminal commands.
Security fixes for message handling and displayed content.
1.7.1
Fixed
Launching the app twice no longer opens a duplicate window.
Fixed a startup failure caused by missing packaged dependencies.
1.7.0
New
Focused graph views now surface what matters instead of everything at once, ranking assets, identities, and credentials by relevance to your Rules of Engagement. Overview-focused is the new default, with a minimal view available.
The planner now starts and resumes on its own, notifies you when it does, and can be turned off in settings.
Rebuilt findings as the single place results land, with a new findings view, faster backend-driven updates, and duplicate marking.
Rebuilt reporting for quicker, more responsive updates, cleaner layout, concurrent deletes, and a new tasks table.
Improved
Compact observations by default, with the latest observation shown and your own messages clearly distinguished.
Aggression level now clearly controls how hard the agent pushes, never what it’s allowed to do — scope limits always take precedence.
Guardrail instructions have been reworded to remove ambiguity, so rules are followed more reliably.
The planner now spins up more terminal agents in parallel when work allows.
Run history is checkpointed mid-run, so long engagements don’t lose context.
Fewer interrupting notifications, with clearer wording when the planner resumes.
Fixed
Monospace and terminal text now render correctly on all platforms, including packaged macOS builds.
Onboarding settings are saved and persist as expected.
Report links and report deletion work correctly in released builds.
Layout fixes in the findings view.
1.6.0
New
Config loading errors are now surfaced directly in the UI.
Browser channel settings can now be cleared without extra configuration.
Improved
Improved security when launching external processes by sanitizing their environments.
Improved planner responsiveness to relevant terminal step updates.
Fixed
Shell mode changes are now applied correctly.
Autonomous mode now starts automatically and completes plans as expected.
1.5.5
New
Workspaces now sort by recent use.
Documentation is now consolidated in the Help Center, with search.
New overview.
Added PromptCaching capability.
Improved
Updated documentation.
Vendored o200k_base encodings.
Switched to the OpenAI Responses API.
1.5.4
New
Help page added.
Reporting has moved to findings.
Docs
Ubuntu 24.04 preparation guide for Hedgehog.
Fixed
Fixes for Ollama
Other visual inconsistencies and glitches
1.5.3
New
Set endpoint and base url for AWS Bedrock connections for private access to AWS.
Configurable embedding dimensions.
Fixed
Bug fixes and stability improvements.
1.5.2
New
Redesigned scoping experience.
Findings: capture findings directly from terminal agents.
Notification center with error notifications and recurring health checks.
Token usage summaries across both your session and workspace, including cached tokens.
Interrupt, steer and handle approvals for terminal agents directly in the planner view.
New shell tooling and performance improvements for large terminal contexts.
Improved
New planner UI matching the updated planner behavior.
Consolidated navigation: the sidebar is gone and everything now lives in the workspace drawer.
Smoother model onboarding and health checks, with recommended models surfaced first.
Iterate on suggested commands before approving or denying them.
Stronger Rules-of-Engagement enforcement backed by a new policy engine, and more reliable long-running operations.
Fixed
Workspaces now sort by last opened.
Reliable agent shutdown and state persistence.
Slow reports are no longer cancelled prematurely.
Improved badge legibility and contrast.
Various stability fixes across fullscreen mode, builds, and graph rendering.
1.4.0
New
Browser automation: agents can now interact with dynamic web pages through local browsers.
New human approval flow for command reviews.
Fixed
Planner call-to-actions surfaced directly in the terminal.
Smarter context summarization with safer truncation boundaries.
Character drops in terminal output during heavy writes.
1.3.0
New
Onboarding flow for AI providers.
Quick chat actions to launch a terminal agent.
License in-app replacement flow.
Notifications and health checks.
Improved
Smoother planner-to-terminal handoff.
Refreshed graph controls and filtering.
Friendlier scoping experience.
Better defaults for summarization based on the chosen model.
Tighter integration between shell output and the graph.
Fixed
Terminal auto-mode UX.
More reliable file picker, packaging, and updates.
Various smaller stability and contrast fixes.
1.2.2
Fixed
Restored the application menu, resolving copy/paste issues that some users were experiencing.
Docs
Updated product documentation
1.2.1
New
Shell mode switcher in the terminal agent UI.
Improved
Cleaner, faster graph with entity types visible on nodes.
Refreshed overview page that updates when you switch workspaces.
Smoother workspace drawer, including better handling of long agent names.
Fixed
Glitches on the overview page when switching workspaces.
Long agent names breaking the workspace drawer.
Incorrect redirect on startup.