The grants standards as callable AI tools: a typed core library and an MCP server that put CROSS and WALKRI inside Claude Code, Cursor, and any MCP-compatible assistant.
CROSS (Common Reporting Outcome Standards Schema) specifies the obligation architecture of a funding round: what grantees must demonstrate at each payment gate, in one of three obligation modes (build a deliverable, produce a measurable change, or recognize past contribution). WALKRI (Working Architecture for Legible, Knowable, Reliable Instrumentation) specifies the quality of the intake fields that collect that demonstration: every field carries a written criterion intent, an operational definition with qualifying and non-qualifying examples, a justified response type, a specified evidence artifact and access path, and a conformance threshold where external standards are referenced. Both are published CC0 at github.com/CrossWalkri.
Clone this repo and point your MCP client at server.mjs in the root. No build step, no package manager.
{
"mcpServers": {
"cross-walkri": {
"command": "node",
"args": ["/path/to/cross-walkri-tools/server.mjs"]
}
}
}Claude Code reads this from ~/.claude/settings.json; Cursor from .cursor/mcp.json. Any stdio-transport MCP client works.
Alternative: install from JSR
Both packages are published on JSR at jsr.io/@cross-walkri.
npx jsr add @cross-walkri/mcp-server{
"mcpServers": {
"cross-walkri": {
"command": "npx",
"args": ["jsr", "run", "@cross-walkri/mcp-server"]
}
}
}For the core library alone: npx jsr add @cross-walkri/core.
| Tool | What it does | Reach for it when |
|---|---|---|
walkri_audit_field |
Audits a field against WALKRI's five requirements; returns per-criterion gaps and a binary verdict: instrument or label | Before any application form goes live; a label cannot be fixed retroactively |
walkri_generate_field |
Generates a WALKRI-conformant field specification from a plain-language description | Designing a form from scratch |
cross_check_gate |
Returns the full requirements for a CROSS gate by type and obligation mode; optionally audits supplied gate content | Configuring or reviewing a payment gate |
cross_configure_round |
Infers an obligation mode from a program description and recommends a validated gate structure | Setting up a round |
cross_classify_framework |
Classifies an external framework against the 50 CROSS+WALKRI primitives across seven layers | Mapping a new framework into the compatibility corpus |
cross_lookup_lens |
Looks up the five cross-cutting lenses (calibration tier, authority source, cultural-methodological lineage, funder typology, framework scope) | Placing a framework or funder in context |
cross_falsifiability_audit |
Applies the four-element falsifiability test, with the five gate-based types and eight failure modes | Testing whether a claim can actually be checked |
cross_audit_round |
Evaluates whether a described round was run correctly under CROSS+WALKRI; verdict plus prioritized gaps | Post-round review, or due diligence on someone else's round |
Full tool reference
walkri_audit_field. Audits a grant application field against WALKRI's five pre-publication requirements. Returns a per-criterion assessment (pass/fail with gap and fix suggestion) and a binary verdict: instrument (all five pass) or label (any fail). Also returns systemic patterns across failing criteria and a prompt template for revising the field with a language model. Use before publishing any application form; a field that goes live as a label cannot be fixed retroactively.
walkri_generate_field. Generates a WALKRI-conformant field specification from a plain-language description of what you want to measure. Returns a draft structural audit and a fully-formed prompt template ready to pass to a language model to produce the complete specification.
cross_check_gate. Returns the complete list of requirements for a CROSS gate of a given type (entry-specification, completion, continuation) operating in a given obligation mode (build, change, retroactive). Optionally accepts the gate content to run against the requirements and identify what is missing.
cross_configure_round. Accepts a plain-language program description and infers an obligation mode, then recommends a gate structure with evidence scope and evidence strength, validates the configuration for internal consistency, and returns a fully-formed prompt template for generating the complete configuration with a language model.
cross_classify_framework. Accepts a description of an external framework and classifies it against the 50 CROSS+WALKRI primitives across seven layers. Returns which layers the framework addresses, which primitives it exemplifies, which are absent, and a prompt template for generating a formal compatibility statement.
cross_lookup_lens. Looks up a dimension of the CROSS+WALKRI Lenses Framework. Five lenses sit above the primitives as cross-cutting metadata: calibration-tier (five tiers from impressionistic to falsifiable, with current-plus-target output shape), authority-source (where binding force comes from), cultural-methodological-lineage (the tradition the framework descends from, including Western institutional, Islamic finance jurisprudence, Indigenous Treaty partnership, Indigenous-led participatory non-Treaty, traditional without administrative infrastructure, and hybrid), funder-typology (ten kinds of funder including bilateral aid agency, multilateral bank or fund, private foundation, government non-aid, pooled fund, individual donor, DAO or protocol, corporate CSR, faith-based, and settlement-administered), and framework-scope-type (what kind of object the framework is). Returns the lens, its enumerated values, and detection criteria.
cross_falsifiability_audit. Applies the four-element falsifiability test from the Falsifiability Architecture document. A falsifiable claim has a pre-committed claim, a named verifying source outside the claimant's control, a drift detection mechanism, and a disclosure obligation. Returns the four structural elements with structural tests, the five gate-based falsifiability types (entry-gate, progress-gate, completion-gate, continuation-gate, portfolio-level), and the eight failure modes (transcendence-claim, declaration-exploit, precision-facade, partial-instantiation, direction-without-destination, vocabulary-without-architecture, correct-map-wrong-territory, frozen-map).
cross_audit_round. Evaluates whether a described grant round was run correctly under CROSS+WALKRI. Checks for entry specification gate, explicit obligation mode, completion gate, WALKRI-conformant field characteristics, and financial accountability. Returns a conformance verdict (conformant/partial/non-conformant) and a prioritized gap list, plus a prompt template for comprehensive AI-assisted evaluation.
packages/core (@cross-walkri/core): TypeScript types and Zod schemas for every CROSS+WALKRI data structure; the WALKRI audit engine (auditField); the CROSS gate logic (getGateRequirements, validateRoundConfig, classifyObligationMode); all 50 primitives as a typed constant array (Primitives Foundation v0.1.7), searchable by name, layer, or keyword; and prompt templates as functions. Zero runtime dependencies, pure TypeScript.
packages/mcp-server (@cross-walkri/mcp-server): wraps the core package and exposes the eight tools via stdio transport.
Core package exports
Types: WalkriField, WalkriCriterion, WalkriAuditResult, WalkriVerdict, CrossRound, CrossGate, CrossObligationMode, CrossGateType, CrossEvidenceScope, CrossEvidenceStrength, CrossPrimitive, PrimitiveLayer
Zod schemas: walkriFieldSchema, walkriCriterionSchema, walkriAuditResultSchema, crossGateSchema, crossRoundSchema (and all constituent schemas)
WALKRI audit: auditField(field: WalkriField): WalkriAuditResult
CROSS gate logic: getGateRequirements(gateType, obligationMode): string[], validateRoundConfig(round: CrossRound): { valid: boolean; gaps: string[] }, classifyObligationMode(description: string): CrossObligationMode
Primitives: PRIMITIVES (typed const array of all 50), getPrimitiveByName(name: string), getPrimitivesByLayer(layer: PrimitiveLayer), searchPrimitives(keyword: string)
Lenses Framework: LENSES (typed const array of the five lenses), getLens(id), getLensValue(lensId, valueId), getAllLensIds()
Falsifiability Architecture: FALSIFIABILITY_ELEMENTS (the four elements), FALSIFIABILITY_TYPES (the five gate-based types), FALSIFIABILITY_FAILURE_MODES (the eight failure modes), getFalsifiabilityType(id), getFalsifiabilityFailureMode(id)
Prompt templates: auditFieldPrompt(field), configureRoundPrompt(description, programType), classifyFrameworkPrompt(description), evaluateRoundPrompt(description)
These tools implement the grants standards: CROSS, WALKRI, and GRAIN, with CRAFT and ORE beneath them as the evidence layer. A sibling repository serves the Coordination Structural Integrity Suite and the Frame Language vocabulary discipline as MCP servers of the same shape: coordination-structural-integrity-suite/tools. The standards' public home, with a working round configurator and prompt library, is crosswalkri.com.
This is a pnpm workspace monorepo.
pnpm install # install dependencies
pnpm build # build all packages
pnpm typecheck # type-check all packagesBoth packages are dedicated to the public domain under Creative Commons Zero v1.0 Universal (CC0). See github.com/CrossWalkri for the canonical standard documents and the compatibility statement corpus. Any grants ecosystem can adopt CROSS+WALKRI without licensing, attribution, or arrangements.