psr-study-assistant-mcp 0.3.1

MCP server that exposes PSR tools to Claude Code

psr-study-assistant-mcp

An MCP server that exposes 55 PSR/SDDP power-system tools to Claude Code through a single unified server. Includes a skill that gives Claude a complete workflow guide — no tool discovery step required.

---

Prerequisites

Requirement	Notes
Python 3.10+	Must be on your PATH
PSR Factory ≥ 5.0.0b111	Licence must be installed and active on this machine
Claude Code	Desktop app or CLI (latest version)

---

Installation

1 — Install the package

pip install psr-study-assistant-mcp

This installs the psr-study-assistant-mcp command on your PATH.

2 — Register the MCP server

Add the following entry to ~/.claude/settings.json under "mcpServers":

"mcpServers": {
  "psr-study-assistant-mcp": {
    "command": "psr-study-assistant-mcp"
  }
}

Claude Desktop uses %APPDATA%\Claude\claude_desktop_config.json instead of ~/.claude/settings.json. The JSON structure is identical — add the block under "mcpServers".

3 - Skill

You can add the skill direclty on claude.

---

Email Notifications (optional)

The server can send email notifications (e.g. when a long simulation finishes). This feature is entirely optional — the server works normally without it.

To enable, add your SMTP credentials to the "env" block of the MCP server entry in settings.json:

"mcpServers": {
  "psr-study-assistant-mcp": {
    "command": "psr-study-assistant-mcp",
    "env": {
      "SMTP_USERNAME": "[e-mail]",
      "SMTP_PASSWORD": "xxxx xxxx xxxx xxxx"
    }
  }
}

The server reads these variables at startup and pre-configures SMTP automatically. You can verify it worked by calling the get_smtp_status tool — it will show your address without requiring a configure_smtp call.

All available variables (only SMTP_USERNAME and SMTP_PASSWORD are required to activate email):

Variable	Default	Description
SMTP_USERNAME	—	Sender email address
SMTP_PASSWORD	—	App password or SMTP password
SMTP_HOST	smtp.gmail.com	SMTP server hostname
SMTP_PORT	587	SMTP port
SMTP_USE_TLS	true	Set to "false" to disable STARTTLS

Gmail — generating an App Password

Gmail requires an App Password (a 16-character one-time code) instead of your regular account password. App Passwords are only available when 2-Step Verification is active.

Step 1 — Enable 2-Step Verification (skip if already on)

1. Go to myaccount.google.com/security.
2. Under "How you sign in to Google", click 2-Step Verification and follow the steps.

Step 2 — Generate the App Password

1. Go to myaccount.google.com/apppasswords.
2. In the "App name" field type PSR Study Assistant and click Create.
3. Google shows a 16-character code (format: xxxx xxxx xxxx xxxx).
   Copy it immediately — it will not be shown again.

Step 3 — Add to settings.json

Paste the values into the "env" block shown above. Spaces in the password are optional — Google shows them for readability but they are not part of the credential.

Revoking the App Password

Go to myaccount.google.com/apppasswords, find PSR Study Assistant, and click the trash icon. Then remove the "env" block from settings.json.

Using a different email provider

Override the defaults in "env":

"env": {
  "SMTP_HOST": "smtp.office365.com",
  "SMTP_PORT": "587",
  "SMTP_USERNAME": "[e-mail]",
  "SMTP_PASSWORD": "your-password",
  "SMTP_USE_TLS": "true"
}

Troubleshooting

Error	Likely cause	Fix
SMTPAuthenticationError	Wrong password or 2FA not enabled	Re-generate the App Password
SMTPConnectError	Firewall blocking port 587	Set SMTP_USE_TLS to "false" and SMTP_PORT to "465"
SMTP not configured	Variables missing or misspelled	Check get_smtp_status — confirm key names in "env"

---

Using the Plugin

Starting a session

Invoke the skill to load the full workflow guide:

/psr-gateway-complete C:/studies/my_case

You can also describe what you want to do:

/psr-gateway-complete C:/studies/my_case  scale thermal marginal costs by 10%

Without the skill

All MCP tools are directly callable from any Claude Code session — the skill just provides additional workflow context.

Setup calls

Every session requires one setup call before domain tools:

Goal	Setup tool
Read / query input data	setup_input_study(path)
Edit a case (safe copy)	setup_edit_study(original, target)
Edit in-place (needs user approval)	setup_edit_study_inplace(path)
Manage output files	setup_output_case(path)

Example — query thermal plant capacity

setup_input_study("C:/studies/case")
get_objects_by_type("ThermalPlant")
get_objects_table("ThermalPlant", ["InstalledCapacity", "MarginalCost"])

Example — scale all thermal costs by 10%

setup_edit_study("C:/studies/case", "C:/studies/case_+10pct")
get_objects_table("ThermalPlant", ["MarginalCost"])
scale_dataframe("ThermalPlant Gas1 [3]", "MarginalCost", 10)
# ... repeat for each plant key
save_study("C:/studies/case_+10pct")

---

Tool Reference

All 55 tools across 10 categories:

Category	Tools
Setup	setup_input_study, setup_edit_study, setup_edit_study_inplace, setup_output_case, save_study
Run	get_case_version, list_available_versions, run_psr_model, get_run_status, read_run_log, kill_run
Discovery	discover_objects, discover_object_properties, get_all_objects, get_objects_by_type, get_neighbors
Query	get_static_properties, get_dynamic_property, get_objects_table
Filter & Aggregate	query_by_property_condition, find_by_reference, count_by_reference, sum_property_by_reference
Input DataFrame	input_get_dataframe, input_aggregate_dataframe, input_filter_dataframe_by_index, input_filter_dataframe_by_value
Binary DataFrame	input_get_binary_dataframe, input_aggregate_binary_dataframe, input_filter_binary_dataframe_by_index, input_filter_binary_dataframe_by_value
Edit	modify_static_element, rename_element, modify_element_code, modify_element_key, get_dataframe_schema, set_dataframe, scale_dataframe, create_element, modify_study_setting
Output Management	list_enabled_outputs, list_enabled_outputs_paths, get_output_num, convert_output, change_output_availability
Output DataFrame	df_info, df_aggregate, df_select_columns, df_combine_columns, df_filter_by_value, df_filter_between, df_filter_by_index, df_merge, df_concat

For full descriptions and usage examples see TOOLS_GUIDE.md and the skill at skills/psr-gateway-complete/SKILL.md.

---

Repository Structure

PSR-Plugin/
├── pyproject.toml                      # Package: psr-study-assistant-mcp
├── skills/
│   └── psr-gateway-complete/
│       └── SKILL.md                    # Analyst workflow guide (55 tools)
└── src/psr_mcp/
    ├── common.py                       # Study setup + artifact discovery
    ├── input_discover.py               # Query and filter tools
    ├── edit_case.py                    # Edit tools
    ├── outputs.py                      # Output management + DataFrame tools
    ├── execution.py                    # Run/execution tools
    ├── notification.py                 # Email notification tools
    ├── state.py                        # Shared study state
    └── gateway/
        └── server.py                   # Unified gateway — all tools registered here

---

How the Gateway Works

All tools are registered at startup and immediately callable — no discovery or activation step is needed.

find_tools(query) is an optional helper that ranks the catalog by relevance to a natural-language query. Use it only when you don't know a tool's name.

---

Adding a New Tool

1 — Write the function in the appropriate module

# src/psr_mcp/input_discover.py
def my_new_tool(obj_key: str, value: float) -> str:
    """One-line summary."""
    ...

2 — Register it in the gateway catalog

# src/psr_mcp/gateway/server.py — CATALOG dict
"my_new_tool": {
    "fn":          _input.my_new_tool,
    "stub":        "Short label (5–10 words)",
    "description": "One-line description shown in find_tools results.",
    "tags":        ["query", "input"],
},

Available tags: setup, input, edit, output, discovery, query, filter, aggregate, dataframe, binary, run.