Skip to content
Agentmart
Submit a tool

Intent-Verified Development (IVD)

33 tools that make AI write, implement, and verify intent against explicit, testable constraints.

Connector by leocelis · GitHub account verified by the MCP Registry

IndexedMITRemote (HTTP)
Write a review

Install

Works with Claude Code, VS Code, Cursor, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install connector/leocelis-ivd

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Remote endpoint

  • https://mcp.ivdframework.dev/mcp Remote (HTTP)

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

33 tools that make AI write, implement, and verify intent against explicit, testable constraints.

Reviews

Write a review

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this server can do and what it asks for, from its tool list and install config. Checked Oct 25, 2026.

Tools
17 read · 5 write · 0 delete · 11 not yet classified
Secrets
1 (API keys, tokens or passwords)
Install scripts
Nothing to install (hosted)
Runs as
Hosted (remote endpoint)

Tools by what they do 33

Read from the source code; the running server may list different tools.

Read 17 · Look things up without changing anything.

  • canon_check
  • canon_check_rules_installed
  • canon_diff
  • ivd_check_placement
  • ivd_discover_goal
  • ivd_find_artifacts
  • ivd_get_context
  • ivd_judgment_capture
  • ivd_judgment_check_installed
  • ivd_judgment_codify
  • ivd_judgment_detect_patterns
  • ivd_judgment_inject_context
Show 5 more tools
  • ivd_judgment_pair
  • ivd_list_features
  • ivd_list_recipes
  • ivd_search
  • ivd_validate

Write 5 · Create, change, send or run things.

  • ivd_import_spec
  • ivd_judgment_propose_recommendation
  • ivd_judgment_save_codified
  • ivd_run_constraint_tests
  • ivd_scaffold

Not yet classified 11 · We could not tell from the name and description.

  • canon_render
  • ivd_assess_coverage
  • ivd_attest
  • ivd_init
  • ivd_judgment_init
  • ivd_judgment_resolve
  • ivd_load_recipe
  • ivd_load_template
  • ivd_propose_inversions
  • ivd_review_intent
  • ivd_teach_concept

Secrets and settings it asks for

Secrets: values you paste in that let it act as you.

  • Authorization

Network hosts it talks to 1

  • mcp.ivdframework.dev · found in remote endpoint, install command

Files it touches

  • Any path the agent passes to its tools

Tools (33)

Read from the source code; the running server may differ. Descriptions are the server’s own words; hidden characters are shown as [U+…].

  • canon_check

    Canon — audit text or a CanonDocument against R-invariants (engine v0.2.0 enforces R1, R2, R5, R10, R13, R14; R13 is the stakes-adaptive format heuristic added in v0.2.0; R3/R4/R6/R7/R8/R9/R11/R12 remain 'partial' by design — see canon_layer.md and audit.py). Returns an AuditReport with per-R findings, severities, an overall verdict in {pass, fail, safety_fail, partial}, and a reproducible hash for R9 diffability. Reference: ivd/canon_layer.md §Audit.

  • canon_check_rules_installed

    Canon — detect whether the IVD + Canon Phase 0a rules block (`<BEGIN-CANON v1.0>` / `<END-CANON v1.0>`) is installed in the active project's agent instruction files (.cursorrules, .clinerules, CLAUDE.md, .github/instructions/canon.md, AGENTS.md, .windsurf/rules/canon.md, optional ~/CLAUDE.md). NEVER writes to disk — when blocks are missing, returns ready-to-paste install_payload entries the AGENT must offer to the user with EXPLICIT permission before editing files. Reference: ivd/canon_layer.md, ivd/recipes/canon-rules.yaml.

  • canon_diff

    Canon — diff two AuditReports (before / after) and return per-R movement (fixed, regressed, unchanged) plus before/after hashes. Use to verify that an edit improved Canon compliance without regressing other invariants. Reference: ivd/canon_layer.md §Audit diff.

  • canon_render

    Canon — render any AI text as a CanonDocument (Setting Phase, confidence-marked body, verification beats, folk-theory notes, identity statement). Tier 1 from raw `text`; Tier 2 from a structured `contract`. Returns markdown by default plus the audit verdict so the caller can gate downstream actions on it. Reference: ivd/canon_layer.md.

  • ivd_assess_coverage

    Assess intent coverage across a project. Scans project structure (directories with code) and compares against existing *_intent.yaml artifacts. Returns coverage report with covered/uncovered modules, coverage %, and prioritized suggestions. The AI agent uses this data to recommend where to add intents.

  • ivd_attest

    Process-attestation gate: check that the agent actually FOLLOWED the IVD method, not just that the artifact is well-formed. Takes the intent plus a structured attestation and mechanically checks Rule 1 (segmentation mandatory at 3+ constraints), Rule 2 (re-read from disk, per-constraint coverage, joint satisfaction), Rule 3 (provenance cross-check — an ai_generated-only test cannot be attested PASS; this is artifact-derived and overrides the agent's claim), and Rule 4 probe coverage. Returns COMPLIANT | INCOMPLETE | VIOLATION. Structure-only: no LLM, no network, no disk write.

  • ivd_check_placement

    Validate that an IVD artifact is in the correct canonical location per framework (co-locate with code).

  • ivd_discover_goal

    When the user doesn't know what to ask: propose 2-3 candidate goals or patterns. Use before 'describe -> AI writes intent' (Principle 6 discovery extension).

  • ivd_find_artifacts

    Discover IVD intent artifacts in a repo. scope=workflow searches workflows/ only; workflow intents alongside coordinator are found with scope=all.

  • ivd_get_context

    Get complete IVD context for AI agent - principles, when to use, available resources. Reduces token usage vs reading all files.

  • ivd_import_spec

    Parse a GitHub Spec Kit or OpenSpec artifact (spec.md) into a constraint scaffold. Read-only, no LLM call. Extracts User Story / Requirement + Given/When/Then scenarios and returns agent_instructions for drafting IVD constraints with a real test: field — the binding neither source format provides.

  • ivd_init

    Initialize IVD in an existing project (brownfield). Creates system_intent.yaml at project root with project context. Use this FIRST when adopting IVD in a project with existing code/docs.

  • ivd_judgment_capture

    Capture a raw correction in < 30 seconds. Writes a ledger entry in state=raw under `.judgment/ledger/raw/`. Use this the moment a real-world correction surfaces (your review, audience signal, runtime error, etc.). Follow up with ivd_judgment_codify. Dormant unless `.judgment/` exists. (IVD v3.0)

  • ivd_judgment_check_installed

    Detect Judgment-phase activation across one or more projects. Two modes: (a) workspace_root set — scan recursively (up to max_depth) for `.judgment/` folders and return per-project summaries (engine_version, ivd_version, baseline / ledger / pattern / recommendation counts, domains); (b) workspace_root unset — report on the single project_root. NEVER writes to disk — read-only by design (analog to canon_check_rules_installed's never-writes invariant). Use this to ask 'which projects in my workspace have the Judgment phase activated?' without spelunking the filesystem. (IVD v3.1, R6)

  • ivd_judgment_codify

    Returns a structured codify prompt for the agent to fill in, plus the original raw correction. The agent fills the 5 codified fields (expected_result, detected_via, diagnosed_cause, proposed_fix, fix_action_type) and then calls ivd_judgment_save_codified to persist. Dormant unless `.judgment/` exists. (IVD v3.0)

Show 18 more tools
  • ivd_judgment_detect_patterns

    Cluster codified ledger entries by (domain, normalized diagnosed_cause). Promotes any cluster with 3+ members (configurable via min_members) to a pattern file under `.judgment/patterns/`. Computes weighted_confidence using domain_depth and freshness using the domain baseline's pattern_half_life_days. Dormant unless `.judgment/` exists. (IVD v3.0)

  • ivd_judgment_init

    Bootstrap the IVD Judgment phase for a project. Creates `<project_root>/.judgment/` with subfolders (baselines, ledger/{raw,codified,paired,resolved,archived}, patterns, recommendations) and seeds per-domain baseline.yaml files. After this runs, the other 9 ivd_judgment_* tools become active. (IVD v3.0)

  • ivd_judgment_inject_context

    Return prioritized Judgment context for downstream agents. Four layers: (1) distilled patterns (active, freshness ∈ {fresh, aging}, sorted by weighted_confidence × member_count), (2) recent codified corrections in the same domain (last 5), (3) ruled_out — REJECTED comparison-pair hypotheses (do-not-retry veto), (4) What Works hypotheses (only corroborated comparison pairs). Soft token_budget cap. Use this to inject judgment knowledge into the next agent run's system message. Dormant unless `.judgment/` exists. (IVD v3.0)

  • ivd_judgment_pair

    Capture a comparison_pair entry — the IVD Judgment alternative to A/B testing for real-world data (Pearl Rung-1: association, not intervention). Each diagnostic_hypothesis MUST list at least one competing_hypothesis. New pairs start at injection_status=plausible; need 2+ independent corroborating pairs to be promoted to corroborated and join the What Works injection layer. Dormant unless `.judgment/` exists. (IVD v3.0)

  • ivd_judgment_propose_recommendation

    Draft a recommendation against a detected pattern. For prompt_patch and intent_revision fix_action_types, also emits a draft_recipe_yaml the agent can review. For capability_addition, infers capability_subtype (build|buy|hire|partner) from the pattern's members. Recommendation lands in approval=pending_user and awaits your gate. Dormant unless `.judgment/` exists. (IVD v3.0)

  • ivd_judgment_resolve

    Close the loop on a ledger entry: record what happened after the fix was applied and whether it held, then transition the entry codified|paired → resolved. This is the outcome-logging step — without it, a future run re-derives a diagnosis that was already settled. Writes a `resolution` block ({outcome, held?, fix_applied?, resolved_at}) and moves the entry to `.judgment/ledger/resolved/`. Dormant unless `.judgment/` exists. (IVD v3.1)

  • ivd_judgment_save_codified

    Persist the agent-filled codified fields. Validates the 5 required fields and capability_subtype (when fix_action_type=capability_addition), then transitions the ledger entry from raw → codified. Dormant unless `.judgment/` exists. (IVD v3.0)

  • ivd_list_features

    Derive feature inventory from all IVD intent artifacts. Use before adding a feature to avoid duplication (large projects).

  • ivd_list_recipes

    List all available IVD recipes with descriptions and use cases. Use this to discover which recipe fits your need.

  • ivd_load_recipe

    Load a specific IVD recipe by name (agent-classifier, workflow-orchestration, etc.). Recipes are proven patterns to use as templates.

  • ivd_load_template

    Load an IVD artifact template (intent, recipe, task, workflow). Use when no recipe matches your need.

  • ivd_propose_inversions

    Scaffold inversion opportunities for a problem (Principle 8: Innovation through Inversion). Use when designing a new feature that has a conventional approach.

  • ivd_review_intent

    Build a human review gate packet for an intent (human review gate). Ranks constraints by risk, surfaces verification.test_cases as worked examples, lists GUESSED constraints pending sign-off. Structure-only — does not auto-approve.

  • ivd_run_constraint_tests

    Opt-in pytest runner for intent-linked tests (local/CI only). Runs ONLY pytest node ids declared in constraints[].test and constraint_satisfiability.joint_satisfaction_test. Disabled unless IVD_TEST_RUNNER_ENABLED=true. Never runs inside ivd_validate.

  • ivd_scaffold

    Create a new IVD intent artifact in the correct canonical location (co-locate with implementation). Works for code AND non-code artifacts (documentation, research, books). For standalone docs (runbooks, specs, guides), use module_path to set docs/ location. Before scaffolding: use task-level only for critical functions (10-20%). For non-default repos, pass project_root.

  • ivd_search

    Semantic search across IVD framework docs. Use when you need specific IVD guidance on a topic without reading all files. Returns relevant chunks with sources.

  • ivd_teach_concept

    When the user lacks technical knowledge: create structured educational artifact explaining a concept they need before IVD flow. Returns YAML with verification questions.

  • ivd_validate

    Validate an IVD artifact before finalizing. Checks structure, required sections, and basic compliance. v3.0 adds Judgment-phase artifact types (baseline, ledger_entry, comparison_pair, pattern).

Signals

Stars
0
Forks
0
Open issues
0
Last commit
Sep 30, 2026 (28 days ago)
Last release
Jul 22, 2026 (3 months ago)

From GitHub (fetched Oct 24, 2026).

Details

Registry name
io.github.leocelis/ivd
Version
3.1.2
Auth
Bearer token
License
MIT
Repository
github.com/leocelis/ivd
Website
ivdframework.dev
First seen
Oct 24, 2026
Last checked
Oct 24, 2026
Source
Official MCP Registrygithub-graphql

More from leocelis

Cite this page

Intent-Verified Development (IVD) MCP server. Agent Mart. https://agentmart.dev/connectors/leocelis-ivd (last updated 2026-10-24)