PERT
Program Evaluation & Review Technique
Weighted three-point estimates with critical-path detection. Surfaces float, slack, and which task to actually unblock.
OSS skill · mcp · api
v0.4 · alpha
From a management unit
to an intelligence organisation.
Claude logic that semi-automates the PMO. Decision tooling for consulting firms. An open-core product that evolves administrative work into intelligence activity.
Skills
4
Claude-native
MCP tools
5
stdio · on PyPI
APIs
REST
FastAPI
Methods
PERT · MC · Bayes
peer-reviewed
§01 / premise
Semi-automate the administrative work with Claude. The PMO of the future becomes the organisation that supplies the insight and direction shaping decision quality.
Not just automating spreadsheets — building a data foundation as your organisation's shared knowledge. As that knowledge updates from real delivery, the agenda your steering committee works through gets sharper.
An AI chat hands you different code and a different answer every time, and a human has to vet each one. Just as AI agents run autonomously inside a harness, your PMO team's Claude uses fixed, maths-based code as its tool — drawing different insight from the same logic as the data updates day by day.
§02 / methods
Vetted logic, turned into tools Claude can use.
Tighter estimates to ease project slippage, realistic sprint lengths derived from your historical data, and near-term risk simulated from a snapshot of where things stand today.
Monte Carlo
Schedule simulation across thousands of runs
Replaces single-point dates with distributions. p50 / p90 ship dates, sensitivity per task, and confidence intervals you can defend.
OSS deterministic seed
Bayesian updating
Estimates that learn from delivery
Updates prior beliefs as tickets close. Replaces the "gut feel re-baseline" with an auditable likelihood update.
OSS conjugate priors
TCO
Total cost of ownership, decomposed
License + ops + integration + training + the hidden line items. Used by SIers to defend procurement decisions.
OSS + industry datasets
+ earned-value, throughput accounting, queueing models, risk-adjusted NPV — see /docs
§03 / install
Easy to set up on Claude, Cursor, or any MCP client.
Install the Skills into Claude or Claude Code. The MCP server runs locally — no account needed, and no data sent to or stored on pmo.run's servers.
# clone the open-source repo $ git clone https://github.com/lemur47/logic $ cd logic # install the skills for Claude Code $ cp -r skills/* ~/.claude/skills/ # Claude can now reason with PERT claude> what's the critical path on the Q3 plan? → skill: pmo/pert critical: A → B → D → F → G (14.2d) slack: C: 2.1d, E: 0.8d
# run the published server — no clone needed $ uvx pmorun-mcp@0.1.1 # or one line into Claude Code $ claude mcp add pmo-logic -- uvx pmorun-mcp@0.1.1 # or wire it into any MCP client { "mcpServers": { "pmo-logic": { "command": "uvx", "args": ["pmorun-mcp@0.1.1"] } } } # pinned to @0.1.1 · on PyPI · 5 tools over stdio
§04 / open core
An architecture you can extend.
The core is open source — drop it into your own environment. Extensions such as plugins are paid, or built by your team; product rollout and consulting are arranged under separate contract.
| Layer | What's in it | Who it's for | License | Pricing |
|---|---|---|---|---|
| OSS Core | Skills, MCP servers, APIs. PERT, Monte Carlo, Bayes, EVM, TCO, queueing. | PMO communities, contract PMOs. | MIT | free |
| PAID Plugins | Proprietary connectors: Jira Cloud, ServiceNow, SAP, MS Project, Salesforce. Industry datasets, benchmark libraries. | Systems integrators, internal platform teams. | commercial | seat / usage |
| SERVICE Strategic | Bespoke analysis, model calibration, rollout, executive briefings, training programmes. | Consulting firms, large transformation programmes. | contract | retainer |
§05 / for whom
Make your PMO work intelligence-led, starting today.
Hand the status reports, plan-vs-actuals, and risk analysis you used to calculate by hand over to Claude. Install the local MCP server into Claude Code or Claude Desktop, connect your spreadsheets or Airtable, then analyse it conversationally.
PMO
Pull plan-vs-actuals from the WBS and run PERT. Feed the results through the Monte Carlo tool, then use Claude as a sounding board for the risk analysis and recommendations.
SIer
Connect to your client's Jira or GitHub and analyse sprints and issues. Up and running the moment it's installed on a loaned laptop. OEM it into your own offering if you need to.
consulting firm
Leave the data analysis to Claude and focus on strategy and change management. Mine your own data platform's history for insight, and build proprietary features as plugins.
§06 / start