BAW Knowledge MCP
A Model Context Protocol server that gives AI agents working knowledge of IBM Business Automation Workflow (BPM 8.6 → BAW 26) and Cloud Pak for Business Automation: how to design, build and test process applications artifact by artifact, engine and REST behaviour (v1, v2, /ops), TWX package structure, coach and UI rules, CP4BA and OpenShift setup, install runbooks, integrations (FileNet, ODM, App Connect, Case, BAI), the product database schema, reference tooling and 156 code-quality rules.
Endpoint: https://bawmcp.dosvak.com/mcp (Streamable HTTP)
Connect your agent
IBM Bob — add to ~/.bob/settings/mcp.json (or .bob/mcp.json in a workspace):
{ "mcpServers": { "baw-knowledge": { "url": "https://bawmcp.dosvak.com/mcp" } } }Claude Code
claude mcp add --transport http baw-knowledge https://bawmcp.dosvak.com/mcpCursor, VS Code, Windsurf, custom clients — the same JSON block in the client's MCP settings (choose the HTTP / streamable-http transport where asked). No account, no key.
What the agent can call
| Tool | Use it for |
|---|---|
ask | one call, one answer: the best passages for a question with topic ids to cite — "how do I send an e-mail from a service flow", "why does my timer not fire on CP4BA" |
search_knowledge, get_topic, get_section, related | hybrid search (synonyms: CSHS, UCA, EPV, PDW, ICN, CPE …) over topics, rules, REST calls and database tables; read topics or single sections; browse by category (method, authoring, how-to, artifacts, general, products, designs, cp4ba, ocp, install, headless, analyzer) |
search_rules, get_rule | TWX Code Analyzer rules (TCA-*) by text, category, severity — the rules the analyzer applies to exported packages |
rest_endpoint | REST calls of the classic v1 API, v2 and /ops with method, path and parameters |
db_schema | product database tables and columns (BPMDB 8.6.2, basdb CP4BA 25.0.1) for reporting SQL |
authoring_checklist | the delivery checklist for a new or changed process app (traditional, CP4BA or both) |
list_tools, get_tool | reference tooling on request: Python harnesses (direct flow runs, Playwright coach sweeps, imports, design sweeps, CP4BA sign-on), generators, wsadmin / shell scripts, SQL, JS API code |
submit_feedback, suggest_topic | tell us what was missing or wrong; propose new topics — BAW / CP4BA-related subjects only |
Resources baw://topic/{id}, baw://rules, baw://rest, baw://guide; prompts design-process-app and review-twx. Topics drafted but not yet validated are marked as such in every result; verified facts say "verified on <version>".
Example
Ask your agent: "Using the baw-knowledge tools, design a purge console for finished process instances on CP4BA and list the analyzer rules the service flows must respect." The agent will search, read the relevant topics (task visibility, /ops delete semantics, queued operations), pull the REST calls and the rules, and cite the topic ids it used.
Usage data and terms
Calls are logged (timestamp, tool, query text, returned ids, client name / user agent, a hashed client address, feedback) so the knowledge base can be improved and used for training. No credentials or personal data are requested or stored; do not send any in queries. The service is provided as is, free of charge, rate limited, and may change without notice. Content © Dosvak LLC; IBM, Business Automation Workflow and Cloud Pak are trademarks of IBM Corporation — this is an independent service, not affiliated with IBM.