Curate-Ipsum

Curate-Ipsum is an MCP server that orchestrates graph-spectral analysis, belief revision, and automated synthesis to produce strongly typed patches with formal guarantees. You can interact with it through an MCP client to generate patches, reason about test and region data, and verify code changes within a structured, multi-framework workflow.

How to use

You can connect to Curate-Ipsum using an MCP client configured for standard input/output (stdio) wiring. The server exposes a suite of tools organized into groups for testing, belief revision, graph analysis, verification, and synthesis. Your MCP client will send requests to the server and receive structured results such as synthesized patches, region analyses, test histories, and verification outcomes.

How to install

Prerequisites: you need Python installed to use the PyPI option, or you can use a containerized runtime with Docker. You should also have an MCP client capable of stdio connections, such as uv or similar tooling.

# PyPI installation
pip install curate-ipsum

# or with uv's runner
uv pip install curate-ipsum

# Docker (embedding model pre-loaded)
docker pull ghcr.io/egoughnour/curate-ipsum:latest

Configuration and startup notes

Configuration is done via standard MCP client wiring to the server. You can run the container or use a local Python setup depending on your workflow.

Additional setup examples

{
  "mcpServers": {
    "curate-ipsum": {
      "command": "uvx",
      "args": ["curate-ipsum"]
    }
  }
}

Run verification stack with Docker (example)

{
  "mcpServers": {
    "curate-ipsum": {
      "command": "docker",
      "args": ["run", "-i", "--rm", "ghcr.io/egoughnour/curate-ipsum:latest"]
    }
  }
}

Available tools

run_unit_tests

Execute unit tests to validate small, isolated components of the code base.

run_integration_tests

Run integration tests to verify interactions between components and subsystems.

run_mutation_tests

Perform mutation testing to measure the robustness of the codebase against small changes.

get_run_history

Fetch the historical records of test runs and synthesis iterations.

get_region_metrics

Retrieve metrics about analyzed code regions, such as coverage and mutability.

detect_frameworks

Identify frameworks detected within the analyzed code base.

parse_region

Parse and normalize a specific code region for analysis.

check_region_relationship

Assess the relationship between code regions (e.g., dependencies, containment).

create_region

Create a new region model to focus analysis and synthesis.

add_assertion

Add a new assertion to the belief revision theory for synthesis guidance.

contract_assertion

Constrain an assertion to reflect refined understanding or evidence.

revise_theory

Apply AGM-based revision to update beliefs given new evidence.

get_entrenchment

Query the entrenchment ordering for minimal contraction decisions.

list_assertions

List current assertions in the belief revision system.

get_theory_snapshot

Retrieve a snapshot of the current theoretical state.

store_evidence

Store evidence from experiments and tests for provenance tracking.

get_provenance

Query provenance information for past decisions and results.

why_believe

Explain why a belief is considered valid given the evidence.

belief_stability

Assess the stability of beliefs over time.

rollback_to

Rollback the system to a previous state or region.

undo_last_operations

Undo the most recent set of operations executed by the system.

analyze_failure

Analyze a failure to identify root causes and missing guarantees.

list_world_history

List historical events affecting the global state of the system.

extract_call_graph

Extract the call graph from the codebase for analysis.

compute_partitioning

Compute graph partitioning to reveal structural regions.

query_reachability

Query reachability within the graph to determine possible paths.

get_hierarchy

Retrieve hierarchical organization of graph nodes.

find_function_partition

Identify function-level partitions within the graph.

incremental_update

Incrementally update the graph after changes.

persistent_graph_stats

Provide statistics about the persisted graph store.

graph_query

Run graph queries to explore relationships and metrics.

verify_property

Verify a property using a backend such as Z3 or angr.

verify_with_orchestrator

Run verification with CEGAR budget escalation orchestrator.

list_verification_backends

List available verification backends.

synthesize_patch

Synthesize a patch using CEGIS, genetic search, and LLM guidance.

synthesis_status

Query the status of an ongoing or completed synthesis run.

cancel_synthesis

Cancel an active synthesis run.

list_synthesis_runs

List past synthesis runs and their outcomes.

rag_index_nodes

Index nodes for RAG-based retrieval.

rag_search

Search the RAG index for relevant context.

rag_stats

Provide statistics about the RAG index and usage.