MCP Tool Reference¶

Curate-Ipsum exposes 30 tools over the MCP stdio transport, organised into six groups.

Testing¶

Tools for running tests, detecting frameworks, and tracking metrics.

`run_unit_tests`¶

Run unit tests for a project and return a summarised result.

Parameter	Type	Required	Description
`projectId`	string	yes	Project identifier
`commitSha`	string	yes	Current commit SHA
`command`	string	yes	Shell command to execute
`workingDirectory`	string	yes	Project root directory
`regionId`	string	no	Scope to a specific region
`framework`	string	no	Test framework name (default: `generic`)

`run_integration_tests`¶

Same parameters as run_unit_tests. Runs integration tests separately so metrics are tracked independently.

`run_mutation_tests`¶

Run mutation tests and return summarised mutation statistics.

Parameter	Type	Required	Description
`projectId`	string	yes	Project identifier
`commitSha`	string	yes	Current commit SHA
`command`	string	yes	Mutation tool command
`workingDirectory`	string	yes	Project root directory
`regionId`	string	no	Scope to a specific region
`tool`	string	no	Framework override (auto-detected if omitted)
`reportPath`	string	no	Custom report file path

Supported frameworks: Stryker (JS/TS), mutmut (Python), cosmic-ray, poodle, universalmutator.

`get_run_history`¶

Return recent test and mutation runs for a project.

Parameter	Type	Required	Description
`projectId`	string	yes	Project identifier
`regionId`	string	no	Filter to a region
`limit`	int	no	Max results to return

`get_region_metrics`¶

Compute PID-like metrics and mutation score for a specific code region.

Parameter	Type	Required	Description
`projectId`	string	yes	Project identifier
`commitSha`	string	yes	Current commit SHA
`regionId`	string	yes	Region identifier

`detect_frameworks`¶

Detect available mutation testing frameworks and project language.

Parameter	Type	Required	Description
`workingDirectory`	string	yes	Project root directory

`parse_region`, `check_region_relationship`, `create_region`¶

Utilities for working with the hierarchical region model (file:path::class:Name::func:name::lines:1-10).

Belief Revision¶

AGM-compliant belief revision powered by py-brs.

`add_assertion`¶

Add a typed assertion grounded by evidence.

Parameter	Type	Required	Description
`projectId`	string	yes	Project identifier
`assertionType`	string	yes	`type`, `behavior`, `invariant`, or `contract`
`content`	string	yes	Assertion statement
`evidenceId`	string	yes	Grounding evidence ID
`confidence`	float	no	Confidence 0.0–1.0 (default: 0.5)
`regionId`	string	no	Scope to a region

`contract_assertion`¶

Remove an assertion via AGM contraction.

Parameter	Type	Required	Description
`projectId`	string	yes	Project identifier
`nodeId`	string	yes	Assertion node to contract
`strategy`	string	no	`entrenchment`, `minimal`, or `full_cascade`

`revise_theory`¶

Incorporate a new assertion, contracting contradictions first (Levi identity).

`get_entrenchment`¶

Get the entrenchment score (0.0–1.0) for an assertion — how resilient it is to contraction.

`list_assertions`, `get_theory_snapshot`¶

Query the current theory state.

`store_evidence`, `get_provenance`, `why_believe`, `belief_stability`¶

Evidence management and provenance tracing.

Rollback & Failure Analysis¶

`rollback_to`¶

Revert the theory to a prior world state by content-addressable hash.

`undo_last_operations`¶

Walk backward N operations through the provenance DAG.

`analyze_failure`¶

Classify a synthesis failure (type mismatch, overfitting, underfitting, etc.) and suggest which assertions to contract.

`list_world_history`¶

List all historical world states with timestamps and reasons.

Graph-Spectral Analysis¶

Call graph extraction, Fiedler spectral partitioning, Kameda O(1) reachability, and hierarchical decomposition.

`extract_call_graph`¶

Extract and analyse a Python project’s call graph. Returns node count, edge count, SCCs, connected components, and function list. Persists the graph to the configured store.

`compute_partitioning`¶

Fiedler spectral partitioning — recursively bipartitions the call graph using the second eigenvector of the Laplacian. Returns a partition tree with algebraic connectivity values.

Parameter	Type	Required	Description
`workingDirectory`	string	yes	Project root
`min_partition_size`	int	no	Minimum partition size (default: 3)
`max_depth`	int	no	Maximum recursion depth (default: 10)

`query_reachability`¶

O(1) reachability check between two functions using the Kameda index on planar subgraphs with BFS fallback for non-planar edges.

`get_hierarchy`¶

Hierarchical decomposition alternating SCC condensation and Fiedler partitioning.

`find_function_partition`¶

Locate which partition a function belongs to, its siblings, and the partition’s entry/exit points.

`incremental_update`, `persistent_graph_stats`, `graph_query`¶

Persistent graph store operations: incremental re-extraction, statistics, and structured queries (neighbors, reachability, node lookup).

Verification¶

Formal verification via Z3 SMT solving and angr Docker symbolic execution.

`verify_property`¶

Run formal verification on a constraint set.

Parameter	Type	Required	Description
`backend`	string	no	`z3` (default), `angr`, or `mock`
`constraints`	list[str]	no	Z3 constraints in mini-DSL
`timeoutSeconds`	int	no	Budget timeout (default: 30)
`maxStates`	int	no	angr state limit (default: 50000)

`verify_with_orchestrator`¶

CEGAR orchestrator with budget escalation (10s → 30s → 120s). Chains verification attempts with progressively larger budgets.

`list_verification_backends`¶

Enumerate available backends and their capabilities.

Synthesis & RAG¶

CEGIS synthesis loop with LLM seeding, genetic algorithm evolution, and RAG-augmented context retrieval.

`synthesize_patch`¶

The main synthesis entry point. Generates a verified patch to kill a surviving mutant.

Parameter	Type	Required	Description
`projectId`	string	yes	Project identifier
`workingDirectory`	string	yes	Project root
`testCommand`	string	yes	How to run tests
`llmBackend`	string	no	`mock`, `cloud`, or `local`
`maxIterations`	int	no	CEGIS budget (default: 50)
`populationSize`	int	no	Genetic algorithm population (default: 20)

`synthesis_status`, `cancel_synthesis`, `list_synthesis_runs`¶

Monitor and manage synthesis runs.

`rag_index_nodes`¶

Index code nodes into ChromaDB for semantic search.

`rag_search`¶

Semantic search with optional call-graph expansion (vector top-k → neighbor expansion → rerank).

`rag_stats`¶

Vector store statistics.

MCP Tool Reference¶

Testing¶

run_unit_tests¶

run_integration_tests¶

run_mutation_tests¶

get_run_history¶

get_region_metrics¶

detect_frameworks¶

parse_region, check_region_relationship, create_region¶

Belief Revision¶

add_assertion¶

contract_assertion¶

revise_theory¶

get_entrenchment¶

list_assertions, get_theory_snapshot¶

store_evidence, get_provenance, why_believe, belief_stability¶

Rollback & Failure Analysis¶

rollback_to¶

undo_last_operations¶

analyze_failure¶

list_world_history¶