Skip to content

Changelog

What’s new

What changed in NEWWORK Projects, and what comes next.

19 October 2026

Specs that quote their sources, feedback and public roadmaps

Specs written from PDFs, emails, decks and sheets, with every quote checked word for word. A feedback inbox ranked by RICE, and a public roadmap customers vote on.

Specs and their sources

  • Every project has a Specs tab. Add PDFs, Word, PowerPoint, Excel, CSV, emails with their attachments, screenshots and photos as sources, up to 50 MB each. See how it works.
  • Write with AI writes the spec from them. Every requirement quotes its source, the server checks each quote word for word, and quotes it cannot find are left out.
  • Where the sources disagree, the spec says so, quotes both sides and says what it follows. Citations open the source at the page, slide or sheet with the quote highlighted.
  • Agents read a spec and its sources with pm-spec-get.

Feedback and the public roadmap

  • Feedback (Pro) keeps what customers ask for in one inbox: by hand, from a call transcript, Intercom, Zendesk, email or Slack. AI groups it into themes with account value, and Priorities ranks the tickets by RICE.
  • Public roadmap (Business): Planned, In progress and Shipped follow their tickets. Visitors vote and send ideas, and hear once when work starts and ships.

19 October 2026

Nova, the AI teammate

An AI teammate on every plan that answers from your projects and code, prepares changes you confirm, and works with your own rights. Plus forecasts, a risk radar, standups and coding agents.

Nova

  • Press ⌘J anywhere. Nova answers from the tickets, specs, time, roadmap, forecasts and risks you can see, searches the web with sources, and opens pages for you. Meet Nova.
  • It reads and changes the product with your own session: row-level security, grants and triggers decide exactly as in the app. Keys, tokens and passwords are out of its reach.
  • Changes come back as a card, from and to, and run only when you press Confirm. Up to 12 per answer.
  • On Pro and above it reads the code of the repositories assigned to a project, and planners turn a conversation into a plan and tickets.

Plan from facts

  • Delivery forecast (Pro): 5,000 runs over the last 12 weeks give Likely, Safe and Almost sure dates and the chance to make the due date.
  • Risk radar (Pro): every morning at 7:00 it flags work gone quiet, reviews waiting, failing checks, blocked work and slipping releases, and tells the right person once.
  • Plan with AI proposes a sprint up to the team’s velocity, Estimate sizes a ticket from similar finished work, and Your standup is written for you every weekday. Weekly stakeholder updates post to your chat channels (Pro).

Coding agents

  • Assign to a coding agent hands a ticket to a Cursor cloud agent with the brief and the checks. Its pull request links itself and the ticket moves to In review when it is done (Business).

19 October 2026

GitLab, deployments, new importers and SCIM

Code from GitHub, GitLab and Bitbucket moves tickets and shows where it is live. Imports from Linear, GitHub, Asana and ClickUp, a two-way Jira sync, and SCIM for Enterprise.

Code

  • Tickets link to pull requests, branches and commits on GitHub, GitLab and Bitbucket whenever they name the key. Reviews and checks show on the ticket, and three switches per project decide when code moves it. How code moves tickets.
  • Deployments and feature flags: GitHub, GitLab, Vercel or any CI report what shipped where, and LaunchDarkly or any flag service reports flags. Releases list the done tickets that are not in production yet.

Switching

  • Import from Linear, GitHub Issues, Asana or ClickUp, like the Jira import: a preview first, people matched, statuses mapped, and old keys keep linking code.
  • Keep in step with Jira (Business): a two-way sync of titles, descriptions, status, people, points and comments while teams move one at a time.

Enterprise

  • SCIM provisioning for Okta and Microsoft Entra ID adds, deactivates and groups people (Enterprise).
  • The audit log exports as CSV or JSON Lines and feeds your SIEM through the API with an audit-only token (Business).

16 October 2026

Brief Standard 1.1, with testing

Acceptance checks must be results you can observe, every check is a test case, and the AI writes and restructures briefs in the standard's format.

Brief Standard 1.1

  • A check that is a task (“Add an export button”) or a wish (“Export is fast”) no longer counts in full: it earns 15 of the 30 points until it is rewritten. The Brief Checker names each one.
  • Every check is a test case. Agents write a test per check, named after it, and pass the check with that test as evidence. A check ending in “(manual)” is passed by a person; agents can’t. All changes.

In the app

  • Drafts from notes, breakdowns and brainstorm plans are written in the standard’s format, with per-ticket open questions and checks a test can prove. Each draft shows its readiness before you create it.
  • Under a hand-written description that would gain from the format, a quiet line offers Restructure with AI: your text moves into sections, nothing is added. Dismiss it for a ticket or for good. Read more.
  • pm-brief-check lists the checks to rewrite, pm-item-start tells the agent how to test each check, and saving a ticket that isn’t ready says what it still needs.

15 October 2026

Docs, single sign-on and the audit log

Product documentation on every plan, and for Business verified pages, a sign-in rule for your company’s domains and an audit log you can export.

Product documentation

  • Every product has a Docs tab, on every plan: pages in Markdown, in a tree, with search and a preview.
  • On Business, merging a ticket’s last pull request proposes documentation updates. Every change must quote the diff, a person applies it, and the page names the pull request it was checked against. How it works.
  • Agents read the docs with pm-docs-search and pm-doc-get, and on Business propose updates with pm-doc-proposal-submit. The MCP server now has 30 tools.

Single sign-on

  • Admins prove the company’s domains with a DNS record and can require everyone with an address on them to sign in with Google or Microsoft. The database enforces it for the app, the API and MCP, and turning it on ends the agent tokens of the people it covers. Set it up.

Audit log

  • Every change with who made it and what it was before, sign-ins, billing events and support visits. Filter, search and export as CSV; entries are kept for a year and can’t be changed. Read more.

15 October 2026

NEWWORK Projects is open

Workspaces, briefs with readiness, agents over MCP, checks with evidence and verified done, plans with a free tier.

Briefs and readiness

  • Write with AI turns a title and notes into a brief for the ticket’s type, with assumptions to keep or strike and open questions instead of guesses.
  • Every ticket has a readiness score from 0 to 100. All four parts make it Ready for agent, with a robot mark on cards and rows.
  • Acceptance checks are their own list on the ticket. An old “Acceptance criteria” list becomes checks with one click.

Agents

  • Connect any MCP client by URL and personal token, or with the newwork-projects-mcp package. On every plan, never metered.
  • pm-brief-check scores a draft before an agent saves it; pm-item-start warns when a brief isn’t ready.
  • Agent timers and checks are marked via AI agent.

Proof

  • Checks pass with evidence: pull request, commit, test run or link. A person’s tick without evidence is Confirmed, with their name.
  • Done tickets show Verified, or exactly what is missing.

Workspaces and plans

  • Sign up with email, a magic link, Google or Microsoft. Onboarding with an example project, a Jira import and agent setup.
  • Members, guests (free on every plan), invites, export and deletion of a workspace.
  • Free, Pro, Business and Enterprise plans, billed per member with Stripe. Business has a 14-day trial.

Also