UOFastCoder · UOFastMCP · Rocket U2

AI agents that speak
Rocket U2 natively.

UOFastCoder agents and skills connect directly to your UniData or UniVerse database through UOFastMCP — reading DICTs, querying files, extracting SB+ screens, and generating modern code, all via the Model Context Protocol.

uofastcoder + uofastmcp
$uofast-mcp start --account PRODVOL
MCP server · 23 tools exposed · UniData PRODVOL

$uofastcoder run screen-extractor
Agent connected via UOFastMCP
Skill: dict-reader47 screens found
Skill: field-mapper → JSON output
Skill: react-stub-gen → components/

$uofastcoder run modernization-auditor
Files: 184 · BP programs: 312
Flags: 23 → audit/migration-flags.md

UOFastCoderagents + skills
UOFastMCPMCP server for U2
$
Works with
Claude Desktop Cursor VS Code Model Context Protocol Rocket U2 UOPY Python FastMCP

Agents and skills that know your U2 data

UOFastCoder agents run pre-built workflows against your live U2 database. Each agent is composed of focused skills — discrete capabilities that call UOFastMCP tools to read, analyze, and transform your U2 data.

AI client
Claude · Cursor · VS Code
Any MCP-compatible AI assistant invokes UOFastCoder agents through natural language or IDE commands.
UOFastCoder
Agents + Skills
Agents orchestrate skills — each skill calls specific MCP tools to accomplish a focused task.
screen-extractor basic-reviewer modernization-auditor api-stub-generator schema-agent
UOFastMCP
MCP Server for U2
Exposes UniData/UniVerse operations as typed MCP tools that agents call directly.
query_file read_record get_dict_items execute_command read_bp_program
Rocket U2
UniData / UniVerse
Your live database. UOPY connects directly — no middleware, no data copies, no ETL pipeline.

Pre-built agents

screen-extractor Agent

Reads every SB+ screen definition from your U2 account via UOFastMCP. Extracts field positions, labels, help text, validation rules, and process chains into structured JSON ready for a React renderer.

→ screens/CUSTOMER-ENTRY.json
fields: 24 · mv: 3 · flags: 2
calls dict-reader · field-mapper · react-stub-gen
modernization-auditor Agent

Full account audit — files, DICTs, SB+ screens, and BASIC programs. G-codes, PROC chains, and external subroutine calls are flagged with severity ratings. Produces a prioritized migration roadmap.

→ audit/migration-flags.md
23 flags · 8 high · 15 medium
calls list_files · get_dict_items · execute_command
schema-agent Agent

Reads all DICT items across every file in an account and builds a complete JSON schema. Use as context for any downstream agent, API design, or documentation task.

→ schema.json
184 files · 2,847 fields mapped
calls list_files · get_dict_items · read_dict_item
basic-reviewer Agent

Reads BASIC programs from BP files, documents subroutine signatures, COMMON block definitions, and called subroutine chains. Identifies programs that reference external files or hard-coded values.

→ bp-review/UPD.CUST.XREF.md
312 programs · 47 flagged
calls read_bp_program · compile_bp_program
api-stub-generator Agent

Takes any U2 file + DICT and generates FastAPI or Express route stubs with correct field names, types, and multi-value handling. Accelerates the REST layer build for screen modernization projects.

→ api/customers.py
GET · POST · PUT · DELETE routes
calls get_dict_items · query_file · read_record
U2 Coding Assistant IDE Plugin

Embedded in Claude Desktop, Cursor, and VS Code. While you code, the assistant reads live DICTs and queries real records — so field names, types, and relationships are always accurate, not guessed.

→ Inline AI context in your IDE
Real DICT data · not hallucinated
calls get_dict_items · query_with_dict_fields

How skills work inside agents

🧩
Composable skills
Each agent = orchestrated skill chain

Skills are focused, reusable capabilities — each one calls specific UOFastMCP tools and produces a typed output. Agents chain skills together to complete a task. A screen extractor runs dict-readerfield-mapperreact-stub-gen in sequence.

dict-reader field-mapper react-stub-gen React component
🔌
MCP-native data access
Every skill call goes through UOFastMCP

Skills don't connect to U2 directly — they call UOFastMCP tools over the Model Context Protocol. This means the same agent workflow runs whether you're in Claude Desktop, Cursor, or a custom pipeline. The MCP server handles authentication, connection pooling, and error recovery.

skill call UOFastMCP tool UOPY U2 data

The MCP layer that makes agents possible

UOFastMCP is a Model Context Protocol server that wraps Rocket's UOPY library and exposes every U2 operation as a typed, AI-callable tool. Agents don't need to know UOPY syntax — they call tools like get_dict_items or query_file and get structured data back.

View UOFastMCP on GitHub ↗
UOFastMCP 23 tools
list_files
read_record
write_record
delete_record
query_file
query_with_dict_fields
get_dict_items
read_dict_item
write_dict_item
delete_dict_item
execute_command
read_bp_program
write_bp_program
compile_bp_program
add_connection
list_connections
close_connection
select_records
read_record_with_fields
write_record_with_fields
Quickstart

Running agents against your U2 database in minutes

Install UOFastMCP, point it at your account, connect your AI client, and start running agents.

  1. 1Install UOFastMCP: pip install uofastmcp
  2. 2Start the server: uofastmcp start --account /u2/PRODVOL
  3. 3Add the config below to Claude Desktop and restart
  4. 4Install UOFastCoder: pip install uofastcoder
  5. 5Run your first agent: uofastcoder run screen-extractor
claude_desktop_config.json
// ~/Library/Application Support/Claude/claude_desktop_config.json
{
  "mcpServers": {
    "uofastmcp": {
      "command": "python",
      "args": ["-m", "uofastmcp"],
      "env": {
        "U2_ACCOUNT": "/u2/PRODVOL",
        "U2_SERVER_TYPE": "ud",  // ud = UniData, uv = UniVerse
        "U2_HOST": "localhost",
        "U2_PORT": 31438
      }
    }
  }
}

# Then run a UOFastCoder agent:
uofastcoder run screen-extractor --output ./screens
uofastcoder run modernization-auditor --output ./audit
uofastcoder run schema-agent --output ./schema.json
Consulting

U2 Modernization services

The same agents and skills power our consulting engagements — you get faster results because the tooling is already built.

🔍
Modernization Audit

AI-assisted technical assessment of your U2 environment using the modernization-auditor agent — every file, screen, and BASIC program analyzed with a prioritized roadmap.

  • Full file and DICT inventory (schema-agent)
  • SB+ screen extraction (screen-extractor)
  • BASIC program dependency map (basic-reviewer)
  • Migration flags with severity ratings
  • Phased roadmap with effort estimates
Screen Modernization

Extract SB+ screens with the screen-extractor agent and build React components backed by API stubs from api-stub-generator — connected to live U2 data.

  • React components for priority screens
  • FastAPI / MCP gateway layer
  • Field validation ported from SB+
  • Process chain hooks (pre/post-write)
  • Deployment guide and runbook
🏗️
Fractional U2 Architect

Expert advisory access plus full use of the UOFastCoder + UOFastMCP toolchain. Your team gets a U2 specialist and production-grade AI tooling without a full-time hire.

  • Advisory access on demand
  • Architecture and schema review
  • Dev team pairing on U2 + MCP
  • Migration sequencing and risk mgmt
  • Month-to-month, cancel anytime

Let's talk about your U2 environment

Whether you're evaluating UOFastCoder, planning a modernization project, or want to see a live agent demo against a U2 database — reach out. We'll respond within one business day.

✉️
🐙
Send us a message

Tell us about your U2 environment and what you'd like to explore.

Goes directly to tech@rokipark.ai. Reply within one business day.

Message sent!

Thanks for reaching out. We'll get back to you within one business day.