0.76.0
keephq/keep0.76.0Jul 13, 2026by takahirom
AI Summary
This release introduces a visual Scenario Graph to visualize project execution structures, built-in guides for AI coding agents, and enhanced UI search capabilities.
Key Highlights
- New Scenario Graph feature visualizes project execution structures and reusable calls.
- Built-in guides for AI coding agents (Claude Code, Codex) to help them navigate the CLI.
- Scenario search toggle added to the UI panel to filter scenarios by goal and note.
New Features
- Scenario Graph visualization
- AI coding agent guides
- UI scenario search
Full Release Notes
## Scenario Graph You can now see the execution structure of a project at a glance: scenario nodes connected by `dependency` edges, with reusable calls expanded per call site into the steps that actually execute (`with` bindings resolved the same way as at runtime). ```bash arbigent graph ``` prints the graph as Mermaid — paste it into GitHub Markdown or [mermaid.live](https://mermaid.live) to view: ```mermaid graph LR classDef reusable fill:#e8f0fe,stroke:#4285f4 n0["login-flow<br/>Log in and reach the home screen"] n1["premium-content<br/>Play premium content"] n2[["login (user=paid)"]]:::reusable n0 --> n1 n1 -.-> n2 ``` The same graph is available in the UI via the new "Scenario graph" button, as a zoomable and pannable dialog with a "Copy Mermaid" action. ## Built-in Guides for AI Coding Agents The CLI now documents itself, so AI coding agents (Claude Code, Codex, etc.) can operate Arbigent end to end just by being pointed at the `arbigent` command: ```bash # List available guide topics arbigent guide # Print a guide arbigent guide writing-yaml ``` Five topics are built in — `setup`, `writing-yaml`, `inspecting-project`, `running-scenarios`, `debugging-failures` — and they are listed at the end of `arbigent --help` so agents can discover them on their own. Combined with `arbigent scenarios`, `arbigent tags`, and `arbigent graph` (none of which need an AI API key or a device), an agent can inspect a project, edit scenario YAML, run scenarios, and read the results without any hand-written project documentation. For example: > Run `arbigent guide writing-yaml` and follow it to add a scenario that verifies the login flow, then run it with `arbigent run --scenario-ids="login-flow"`. ## Scenario Search in the UI The scenario list panel now has a search toggle that filters scenarios by goal and note (case-insensitive). Ancestors of matching scenarios stay visible so the dependency structure remains readable while filtering. ## Toolchain Upgrade Arbigent is now built with Kotlin 2.3.21, Compose 1.10.3, and Jewel 0.37. ## What's Changed * Add search by goal and note to scenario list panel by @takahirom in https://github.com/takahirom/arbigent/pull/388 * Upgrade to Kotlin 2.3.21, Compose 1.10.3 and Jewel 0.37 by @takahirom in https://github.com/takahirom/arbigent/pull/389 * Add guide command printing agent-facing guides from arbigent CLI by @takahirom in https://github.com/takahirom/arbigent/pull/390 * Update CLI docs in README and document AI agent usage by @takahirom in https://github.com/takahirom/arbigent/pull/391 * Add scenario graph view to CLI and UI by @takahirom in https://github.com/takahirom/arbigent/pull/392 * Extract ComboBoxItem composable and restore item content descriptions by @takahirom in https://github.com/takahirom/arbigent/pull/393 * Share reusable node colors from graph model and consolidate node styling by @takahirom in https://github.com/takahirom/arbigent/pull/394 * Fix non-deterministic scenario graph screenshot by fixing dependency scenario id by @takahirom in https://github.com/takahirom/arbigent/pull/395 **Full Changelog**: https://github.com/takahirom/arbigent/compare/0.75.0...0.76.0