Orangepro MCP server
{
"mcpServers": {
"orangepro-mcp": {
"command": "npx",
"args": [
"-y",
"@orangepro/mcp-server@latest"
],
"env": {
"OPENAI_API_KEY": "sk-...",
"OPENAI_BASE_URL": "sk-...",
"ANTHROPIC_API_KEY": "sk-ant-...",
"OLLAMA_BASE_URL": "https://..."
}
}
}
}{
"mcpServers": {
"orangepro-mcp": {
"command": "npx",
"args": [
"-y",
"@orangepro/mcp-server@latest"
],
"env": {
"OPENAI_API_KEY": "sk-...",
"OPENAI_BASE_URL": "sk-...",
"ANTHROPIC_API_KEY": "sk-ant-...",
"OLLAMA_BASE_URL": "https://..."
}
}
}
}{
"mcpServers": {
"orangepro-mcp": {
"command": "npx",
"args": [
"-y",
"@orangepro/mcp-server@latest"
],
"env": {
"OPENAI_API_KEY": "sk-...",
"OPENAI_BASE_URL": "sk-...",
"ANTHROPIC_API_KEY": "sk-ant-...",
"OLLAMA_BASE_URL": "https://..."
}
}
}
}{
"servers": {
"orangepro-mcp": {
"command": "npx",
"args": [
"-y",
"@orangepro/mcp-server@latest"
],
"env": {
"OPENAI_API_KEY": "${input:openai_api_key}",
"OPENAI_BASE_URL": "${input:openai_base_url}",
"ANTHROPIC_API_KEY": "${input:anthropic_api_key}",
"OLLAMA_BASE_URL": "${input:ollama_base_url}"
}
}
},
"inputs": [
{
"id": "openai_api_key",
"type": "promptString",
"description": "Openai api key",
"password": true
},
{
"id": "openai_base_url",
"type": "promptString",
"description": "Openai base url",
"password": true
},
{
"id": "anthropic_api_key",
"type": "promptString",
"description": "Anthropic api key",
"password": true
},
{
"id": "ollama_base_url",
"type": "promptString",
"description": "Ollama base url",
"password": true
}
]
}{
"mcpServers": {
"orangepro-mcp": {
"command": "npx",
"args": [
"-y",
"@orangepro/mcp-server@latest"
],
"env": {
"OPENAI_API_KEY": "sk-...",
"OPENAI_BASE_URL": "sk-...",
"ANTHROPIC_API_KEY": "sk-ant-...",
"OLLAMA_BASE_URL": "https://..."
}
}
}
}file .mcp.json (in your project root)file claude_desktop_config.jsonfile ~/.cursor/mcp.jsonfile .vscode/mcp.jsonfile ~/.codeium/windsurf/mcp_config.jsoninstall npx -y @orangepro/mcp-serverversion 0.2.40
What it does
OrangePro MCP server maps public behaviors in a codebase, assigns evidence tiers to each (from 'No Signal' to 'Dynamically Proven' via mutation testing), and generates grounded tests for uncovered gaps. It is aimed at developers and coding agents who want to identify untested behaviors and verify test quality through mutation-kill oracles. All analysis runs locally; no source code is uploaded.
- Identify which behaviors in a codebase have no test coverage, weak structural links, or only runtime coverage, ranked by blast radius.
- Generate runnable test drafts (Jest, Vitest, pytest, JUnit, etc.) for the highest-risk uncovered behaviors, scoped to a branch diff or the full repo.
- Run a mutation-kill oracle against a specific behavior to confirm that a test actually fails when the behavior changes, elevating it to 'Dynamically Proven'.
Tools this server exposes
| Tool | Does |
|---|---|
| orangepro_start | Runs analysis, builds the evidence report, and returns recommended next actions in one step. |
| orangepro_analyze_sources | Builds or refreshes the evidence graph from the codebase. |
| orangepro_generate_tests | Generates grounded test drafts for behaviors with weak or missing test evidence. |
| orangepro_prove | Runs the mutation-kill oracle on a specific behavior to confirm a test breaks when the behavior changes. |
| orangepro_prove_loop | Runs setup, dynamic proof, and report refresh for a single behavior in sequence. |
| orangepro_find_test_gaps | Lists behaviors with weak or missing tests, ranked by risk. |
| orangepro_graph_score | Returns the graph readiness score (0–100) for the workspace. |
| orangepro_status | Reports current workspace state without generating or modifying anything. |
| orangepro_doctor | Recommends the next type of evidence to add to improve coverage quality. |
| orangepro_rtm | Produces a requirements traceability matrix linking behaviors to tests. |
| orangepro_stats | Returns aggregate statistics about behaviors, evidence tiers, and gaps. |
| orangepro_changed_impact | Shows which behaviors a diff touches, given a git base reference. |
| orangepro_record_run | Records the result of a test run into the evidence graph. |
| orangepro_explain_test | Explains why a particular test was generated, including its grounding and scenario bucket. |
| orangepro_export_evidence_pack | Exports a metadata-only evidence pack without including source code. |
| orangepro_update_graph | Performs an incremental update to the evidence graph. |
| orangepro_ai_links | Suggests weak behavior-to-symbol links using an optional AI provider. |
| orangepro_ai_flows | Discovers candidate flows across service boundaries using an optional AI provider. |
Before you enable it
An MCP server runs with whatever access you give it. Start with a development account rather than production, and check which tools the server exposes before you let an agent call them unattended.