The tools I wish existed.

Visiban→

Swimlanes, real-time collaboration, dwell-time analytics, and a full audit trail of every card movement. The first MacroDream project to ship — and the source of the governance system now used across the portfolio.

There is also a hosted demo at try.visiban.com — the live app, open to click through.

TruePPM→

Most project tools store dates. TruePPM computes them. At the center is a real Critical Path Method engine with Monte Carlo risk analysis — and the reason to believe that claim is that the engine ships as its own library. pip install trueppm-scheduler and check the math against MS Project in sixty seconds: no server, no account, no trust required. No other PPM tool, open or commercial, makes its scheduling engine auditable in isolation.

Around that engine sits a platform where the Kanban card and the Gantt bar are the same object — not two records kept in sync. Close a sprint and the master schedule reforecasts from measured velocity. MS Project has no agile surface; Jira has no critical path; the open-source field has neither Monte Carlo nor a separable engine. That gap is the thesis.

The Apache 2.0 community tier is a complete P3M practice for a self-hosted team — full CPM engine, Monte Carlo, programs with their own backlogs, offline-first sync, real-time collaboration, time tracking, baselines, and a production Helm chart. No seat count, no feature flags, no SSO tax. MS Project import/export means switching costs an afternoon, not a quarter.

0.4, the first beta, adds what makes it a first-class citizen of an AI-assisted stack: trueppm-mcp, a read-only MCP server published alongside the platform. Point Claude Desktop or any MCP client at your own instance with a personal, expiring, read-only token, and nineteen tools read those computed schedules as first-class facts rather than guessing at them — including a non-mutating what-if, where an agent proposes a scenario, the engine recomputes it, and the answer comes back computed with the real plan untouched. Unlike the engine, it is not a standalone tool: it is a thin adapter over the TruePPM API, and it can see nothing its token's owner could not already see.

Alongside it: single sign-on against your own identity provider, read-only share links so a schedule can travel beyond its own instance, and a read-only mirror of a contributor's own assigned Jira items into one view.

When an organization outgrows a single team, the Enterprise tier adds the governance layer on top: portfolio dashboards, cross-program resource leveling, org identity governance (SAML, SCIM, directory sync), and approval workflows.

And what it is not, stated plainly because a product that only lists its wins isn't being evaluated honestly: there is no resource leveling, one constraint type where MS Project has eight, no cost model before 0.5 and no earned value before 0.8, and a measured ceiling of roughly a thousand tasks in the schedule view. It is not a Primavera P6 replacement and does not claim to be. The gap list is published and maintained.

Next: the first beta was tagged September 15, 2026, and the 0.4 line hardens under further beta tags — leaving beta is a quality judgment, not a date. 0.5 is targeted for Jan 4–18, 2027, counted from the close of the beta period: resource allocation that warns about an over-allocation before you save it, a project status update built from numbers the engine already computes, a labor cost model with a P80 cost beside the P80 date, an installable PWA, and MCP plan mode, where an agent's proposed change comes back with a verdict and its impact and nothing commits.

To see it before installing anything, there is a hosted read-only demo on the bundled sample program. It is a live instance, not a screenshot tour, and nothing you do there is saved.

trueppm-scheduler→

TruePPM's scheduling core as a standalone Apache 2.0 Python library — all four dependency types, calendar-aware lag, multi-calendar composition, cycle detection that names the offending path, critical path, float, and Monte Carlo with per-task sensitivity. No server, no database, just the math the platform runs on, versioned in lockstep with it. A Rust/petgraph twin is held in CI conformance against it.

New in the 0.4 line: derive_value() answers why a computed date or float is what it is, naming the predecessor, dependency type, and lag that won — and the public API is property-fuzzed in CI, so every input either schedules or raises a documented error.

Publishing it separately is a positioning decision, not a packaging convenience. Every scheduling tool claims its math is real; this one can be checked in sixty seconds by someone who does not trust us, and diffed against MS Project. A falsifiable claim is the only kind worth making about scheduling.

Blueprint→

Stack-agnostic project governance template for teams using GitLab and Claude Code. CI pipeline scaffolding, git hooks, changelog workflow, Claude Code agents and skills, and MR conventions — without opinionating on your programming language or framework. The one idea underneath it: a gate's only value is that its green is evidence, so every gate has to prove it still bites.