Changelog

What's new in Invoked.

v0.4.0

Latest
New
  • Datasets and Prove the Lift turn a set of real tasks into a scored benchmark, then run two models head to head and see the measured improvement on your own data. An LLM judge grades every answer with a reason, and the result is an honest number: a regression shows as a negative lift, never a faked win
  • The judge defaults to the cheapest model you have already connected, with a local Ollama judge as a first-class, fully private option. Invoked never hosts a judge; grading runs on your keys or your machine
  • Publish any lift result as a public benchmark page a shareable link anyone can open, with the same animated reveal, the per-row breakdown, and the headline lift. Publishing is opt-in and discloses exactly what becomes public
  • A public benchmark directory at invoked.ai/lift browse and filter shared results by model. Listing your own result is your choice, offered on publish and never required

v0.3.1

New
  • Bonsai support run prism-ml’s new local-first Bonsai models (27B and 8B) inside Invoked, out of the box via Ollama. No API key, nothing leaves your machine. Pick it from a one-click model chip in Settings and drive it through a real multi-agent harness
  • A ready-to-run “Bonsai · Local research swarm” harness ships built in a Planner, two parallel scouts, a guardrail, and a Writer, every agent powered by Bonsai locally
  • Quick-pick model chips under the Model field the popular and newly-supported models for each provider are now one click to select, not just something you type

v0.3.0

New
  • Publish a new step that turns a surface into a live, governed endpoint on your API gateway. Connect your Kong Gateway, preview exactly what will be created, run a dry-run diff, and publish. Invoked pushes the config for you; your services, routes, and API-key protection appear on the gateway, ready to route
  • Export any surface as a portable OpenAPI 3.0 spec download it or copy it, and import into Postman, SwaggerHub, AWS API Gateway, Azure API Management, and most API tooling
  • Your surface stays the source of truth: edit it once, then re-publish across every gateway. Invoked keeps the definition, the credentials stay in your OS keychain, and only the config travels
Improved
  • Deleting a surface that’s shared with your team now warns clearly that it will be removed for everyone, and reliably stays deleted instead of syncing back

v0.2.5

New
  • Share a surface with your team everyone connects to the same tools in one step, and the Dashboard shows that surface’s combined reliability and latency across your whole team. Credentials always stay on your own machine
  • A Workspace switcher in the user menu move between your personal workspace and each team deliberately, and your choice is remembered
Improved
  • Signing in always lands in your personal workspace; joining a team never silently switches what your Dashboard shows
  • The team sync action now shows clear progress a spinner while syncing and a confirmation when it’s done

v0.2.4

Improved
  • Dashboard analytics and in-app notifications now initialize reliably in packaged builds
  • The Dashboard focuses your most recently used surface and shows a clear message when a surface has no activity yet
  • Added an About panel under Settings with version, workspace stats, and last-backup details

v0.2.3

Improved
  • Sign-in reliability improvements across macOS, Windows, and Linux builds

v0.2.2

New
  • Import from LangGraph bring a compiled LangGraph graph onto the harness canvas from its JSON export. Nodes, edges, and conditional branches carry over, with an honest report of what mapped
  • Send feedback and report issues straight to the Invoked team, with diagnostics attached automatically. Team plans can route tickets into their own Linear, Jira, GitHub, and more
Improved
  • Imports now land directly on the Build canvas, focused on the graph’s starting point so complex imports open readable instead of as a distant tangle
  • Import and Export now sit together under Toolkit → Data & Runs your data, in and out, in one place
  • The harness canvas panels slide open and closed smoothly, and the node palette no longer collapses on its own
  • Lighter canvas animations on large graphs, so big imported flows stay responsive
  • Backup now captures workspace changes reliably, with clearer status about when it last ran

v0.2.1

Improved
  • Run & Ship now anchors the bottom of the sidebar, so the build-run-ship flow reads top to bottom
  • The app opens at a comfortable size on large displays and remembers the size you set
  • Zoom controls on the harness canvas for navigating larger flows

v0.2.0

New
  • Run & Ship run a harness live and watch every step, then export portable integration code: React (component + hook), a TypeScript client, Angular, Vue, and Svelte. Bring your own key; it runs anywhere
  • Import hub bring your existing setup in: OpenAI Agent Builder → a harness; OpenAPI/Swagger, Postman, Insomnia, and cURL → a surface of callable tools
  • Typed harness parameters declare typed inputs that render as a form and flow into the generated code’s function signature
  • App-wide drag-and-drop file intake
  • Knock-powered in-app notification bell with unread dot indicator
  • Workspace ID persisted in SQLite migrated automatically from localStorage on first launch
Improved
  • Run & Ship is key-aware: it tells you exactly which key or token to add, never a silent failure
  • Notification bell alignment, popover shadow, and border radius polished

v0.1.5

New
  • Import: a top-level way to bring your existing setup into Invoked (OpenAI Agent Builder, OpenAPI/Swagger, Postman, Insomnia, cURL)
  • Migration microsite for teams leaving OpenAI Agent Builder, served at its own domain
  • Team backup & restore: export your workspace to a local folder, restore from a file, symmetric with cloud backup
  • Guided onboarding walkthrough for first-run
Improved
  • Product analytics wired end to end (web + desktop), consent-gated, no prompts or PII ever leave your machine
  • Per-surface example prompts so each surface suggests the right actions
  • Reliability: in-app App Logs and a cloud-backup toggle

v0.1.2

New
  • Exchange: an in-app directory to search and add surfaces and skills from the public catalog, with a multi-select loadout
  • Surfaces: connect MCP servers, tool sets, and APIs, each fully isolated, and switch between them with a keystroke
  • Harnesses promoted to a first-class section with Build, Evals, and Export
  • Teams: workspace-scoped sharing of surfaces, skills, and harnesses, with team reference runs
Improved
  • Docs rewritten for the current product (Surfaces, Exchange, Harnesses, Tokens, Backup)
  • Sidebar and navigation polish; Exchange promoted in Explore

v0.1.0

Initial beta
New
  • Initial public beta release
  • Runs, Evals, Dashboard, Skills, and Multi-Agent Canvas panels
  • 17+ built-in tools across filesystem, web, shell, git, and Tinybird
  • 5 LLM providers: Anthropic, OpenAI, Groq, Ollama, AWS Bedrock
  • Command palette (⌘K) for instant navigation
  • Local-first all data in SQLite, no account required