Aibvf MCP

What aibvf-mcp does

The aibvf-mcp MCP server turns AI initiative and portfolio inputs into structured planning assessments. It covers four pillars: Strategic Alignment, Financial Return, Change Enablement, and Governance Risk. Results can include a Stop, Fix, or Accelerate verdict, the rule behind that verdict, unresolved evidence, estimated or supplied inputs, a decision score, and a recommended next action.

The model is deterministic for a given engine version and identical inputs. Its scores are rule-based planning outputs, not calibrated probabilities of delivery success. The planning benefit calculation also uses a readiness-adjusted revenue scenario; project costs, margins, timing, and benefit overlap require separate analysis.

How it works

Use assess_ai_initiative when starting from a plain-English proposal. The tool resolves the description, asks for missing decision inputs, and returns an assessment. For explicit structured inputs, score_initiative applies the pillar rules and the work-architecture gate. A gap, partial assessment, or missing evidence about workflow, roles, decision rights, or measures can keep an otherwise favorable proposal at Fix.

Stop conditions include Governance Risk of at least 70 or Financial Return of 20 or less. Accelerate requires Strategic Alignment, Financial Return, and Change Enablement of at least 60, Governance Risk of at most 40, and all four work-architecture checks. These thresholds should be read alongside the supplied evidence and the model's disclosed assumptions.

Portfolio work follows a separate sequence: assemble inputs, validate the portfolio document, score it, and produce rollout waves. Aggregate benefits need review for shared work and overlapping benefits before being treated as a total.

Setup and configuration

The aibvf-mcp MCP server runs locally through an npm command:

{
  "mcpServers": {
    "aibvf": { "command": "npx", "args": ["-y", "aibvf-mcp"] }
  }
}

The README identifies Claude Desktop, Claude Code, and Cursor as local MCP clients. Restart the client after adding the configuration. Windows uses cmd /c around the command. A hosted Streamable HTTP endpoint is also available at https://mcp.aibvf.com/api/mcp, including as a custom connector for claude.ai.

Telemetry is enabled by default. Set AIBVF_TELEMETRY_DISABLE=1 to prevent events and creation of the local install identifier. AIBVF_TELEMETRY_URL and AIBVF_TELEMETRY_KEY can direct telemetry to another backend. The documented events exclude portfolio content, revenue figures, numeric pillar scores, and personal identifiers.

Tools and capabilities

The 13 exposed tools include:

• Assess, score, and improve individual initiatives.
• Assemble, validate, score, and sequence portfolios.
• Diagnose process signals and infer readiness.
• Calculate directional pace-layer drag scenarios.
• Retrieve planning benchmarks and list supported taxonomy values.
• Map everyday business terms to the accepted taxonomy and identify unresolved terms.

recommend_improvements proposes pillar actions and work redesign steps for Fix or Stop results. Proposed actions should be reviewed by the people who own the work, then the initiative should be re-scored after evidence changes.

Limitations and notes

The aibvf-mcp MCP server does not certify regulatory compliance or perform clinical validation based on module labels. External research informs the questions but does not validate the model's rates, multipliers, readiness percentages, or drag rates. Organisations remain responsible for checking the evidence behind supplied values.

Use the results as a documented decision-planning aid. Review the scoring formulas and worked example before using an output in a funding decision. Package versions identify implementation behavior, while the public portfolio specification has its own version.