/00 Changelog

What is new in Arq

  1. Databases as first-class parts of a system

    • The palette offers every symbol of the view's notation, grouped as the notation groups them, plus every other kind; new block shapes for nodes, queues, networks, agents, components and modules. The model page opens on its views, each with Open canvas, and the outline shows the model's purpose, facts, contents, health and versions.
    • The canvas: palette tiles drawn like the blocks they create, a palette you can collapse and resize, right-click menus on the canvas and on blocks, the relationship type chosen where the connection ends, a compact legend and quick keys on the canvas, and the model shown before the view.
    • Git sync: models as text in your repository, changes opened as pull requests, and edits made there brought back for review.
    • Integrations: connect Backstage, Kubernetes, SBOMs, Terraform, OpenLineage or any JSON records by upload, signed webhook or scheduled pull; what they report appears beside the model, and differences arrive in the inbox.
    • Organization console: members across workspaces, verified domains that bring colleagues in, SCIM 2.0 provisioning with groups mapped to workspace roles, and audit retention.
    • Review agents propose classifications and controls into the inbox, in the background if you opt in; CMMC 2.0 Level 2 controls; matrices for activities × capabilities and controls × what they protect.
    • Behaviour views: UML sequence diagrams, UML state machines and C4 dynamic views with numbered steps; derived relationships previewed in the inspector.
    • Your own kinds and relationship types, such as a Claims system that is a kind of Application, with their own attributes; and one-step merge of duplicate elements.
    • Reports: thread coverage with what is missing, the impact of any element in both directions, and the validation report, as Markdown or CSV.
    • Signed webhooks for model changes, proposals, comments and more, and an audit stream to your SIEM; every change to a model is now in the audit log.
    • A REST API at /api/v1 with an OpenAPI description, and new MCP tools: query impact and dependencies, compare baselines, export, and create or lay out views.
    • Comments on a model, its elements and its views, with replies and resolve; a Commenter role for reviewers who discuss but do not edit.
    • Federation: build what Lattix, Savant and Kepler receive from a baseline, see what the receiver would flag and what changed since the last export, and download the payload with a receipt.
    • Import from the tools you already have: Backstage catalogs, OpenAPI and AsyncAPI, CycloneDX and SPDX SBOMs, Terraform state, Kubernetes manifests, OpenLineage events and Open Threat Model files, all reviewed before they land.
    • Data models export as Postgres DDL, DBML and JSON Schema, and import from SQL DDL; any model exports as PlantUML.
    • Offline: a model or view you have opened opens again without a connection; edits made offline sync when you reconnect. Signing out clears the offline copies.
    • Agent oversight: rate runs, see every run in a workspace, and acceptance rates by agent and by kind of change in settings.
    • Views have a legend and a keyboard map: arrows move along relationships, ? shows the legend, / searches the palette, Ctrl+L lays out. The palette gains a Library of hub building blocks.
    • Capability maps, value streams and org charts: new notations whose views lay themselves out from the model.
    • Brief: every model has a structured brief with requirements and attached documents (.md, .txt, .docx, .pdf), and Arq drafts views from it. In a view, Ctrl+K or Cmd+K sends free text to Arq.
    • Inspector: Appears in (models and views that show an element), Trace (up to why it exists, down to what realizes it) and Exports (which of Lattix, Savant and Kepler receive it).
    • Imports are reviewed first: see what a file adds, link elements the workspace already has instead of copying them, and keep the file's hash and importer in history. New: Mermaid (flowchart, class, ER, sequence), Structurizr JSON, and CSV or Excel tables for inventories, capabilities, org charts and RACI.
    • Workspace attributes: add your own attributes to any kind, such as a cost center or a service tier, and every change is checked against them.
    • Branches: update a plan from Current when Current moves on, and freeze a plan as a baseline that stays as it was.
    • Threat modeling: each threat suggests NIST SP 800-53 controls for its STRIDE category; adding one mitigates the threat and protects the exposed element.
    • Personnel: skills needed and held as heatmaps with gaps marked, and staffing as FTE per role and quarter, each downloadable as CSV.
    • Views on a branch: draw the target state on its own layout, starting from Current's. Promoting the branch brings what you drew onto Current, alongside any changes made there meanwhile.
    • Branches: plan a transition, a target or an alternative beside Current, compare it with where it started or with Current now, see which other models it affects, and promote it when the plan becomes real.
    • Security and data flow notation: draw trust boundaries and classified data flows. The Threats tab lists every flow that crosses a boundary, flags those no control protects, suggests STRIDE threats you can add in one click, and downloads a threat report.
    • Matrix editor: assign RACI across roles and activities, or CRUD across components and data, cell by cell. An activity with assignments needs exactly one Accountable.
    • A system or component now contains its databases and other data stores, so a C4 database is a container of its software system. Data stores have a store type alongside technology and classification.
    • Rule packs: built-in data classification, ownership and C4 hygiene packs, or your own rules as JSON, checked in every model; block rules stop baselines.
    • Audit log: organization owners and admins filter what was recorded and export it as CSV or JSON Lines.
    • Baselines: freeze a model under a name, compare any two states as a semantic diff, and export a frozen, hashed state.
    • Editing together: changes to a view reach everyone in about half a second, and the header shows who else has it open.
    • Every element has a stable link that opens it in its home model; copy it from the inspector.
    • Signing in or confirming a new account takes you straight to your workspace.
  2. Structurizr import and validation

    • Import Structurizr DSL workspaces: people, software systems, containers and relationships, with a view for each landscape, context and container view.
    • ArchiMate, C4 and BPMN are now declared at fidelity L2: each round-trips its test fixtures.
    • UML class, component, deployment and use case views (fidelity L1) over the same model.
    • Import and export whole workspaces as .arq bundles, and keep editing views offline.
    • Agent run history per model, with accepted elements linked to the run that proposed them; models have an editable lifecycle.
    • Home, workspace inbox, workspace settings and account pages.
    • Canvas: align and distribute a multi-selection, reshape edges with bendpoints, and reconnect an edge to another block.
    • Validation: a whole-model check with errors, warnings and suggestions, a glyph on affected blocks, and click-to-select. Also available to coding agents as arq_check_model.
  3. Architecture hub, a new look and the documentation

    • Architecture hub: import validated reference architecture into your models, and update without losing your changes. Browse it at /hub.
    • Arq now uses the Lattix visual language: warm greys, fine lines and a drawing grid that moves with the canvas.
    • Documentation, pricing, security, notation and use-case pages.
  4. First release

    • One model across ArchiMate, C4, BPMN and ER views, with an outline editor, a catalog and a text projection.
    • Draft views from a brief, review proposals block by block, and run commands in plain language. Hosted credit or your own key.
    • Import and export ArchiMate Exchange and BPMN 2.0; export Structurizr DSL, Mermaid, SVG and PNG.
    • MCP server for coding agents, and the context pack for your repository.
    • Revisions with a semantic diff, read-only share links, members and invitations.
    • A browser-only demo, a 14-day trial with no card, and one plan at $10 per user per year or $1 per month.