Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@Fusion Schema MCP Server show columns for the FND_USER table"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Standalone HTTP MCP server that grounds a Fusion-SQL agent against the Oracle Fusion schema catalog. The models already know most Oracle ERP names (from EBS on-prem), but Fusion Cloud ≠ EBS — so these tools validate and correct the objects/columns the agent thinks it knows, and surface real join paths (declared FKs + mined relationships).
Scope: grounding only. No SQL parsing, no SQL execution/validation (that needs a live Oracle DB — later round). No agent, no CloudBeaver UI (later rounds).
Tool | Purpose |
| FTS over name + description + module; discovery by description |
| metadata: type, module, description, primary key, column count |
| columns with type/size/nullable/description/PK flag |
| exists? + fuzzy "did-you-mean" (EBS→Fusion drift) |
| per-column exists + suggestions |
| indexes with ordered columns + uniqueness |
| declared FKs + mined relationships, tagged + confidence = join paths |
Related MCP server: OracleDB MCP Server
data/ holds the source metadata (copied from ../../Bip/DB_SCHEMA via
scripts/copy-data.sh): META_TABLES/COLUMNS/INDEXES/PKEYS/FKEYS.csv +
mined_relationships.json. The compile step filters junk table rows to TABLE/VIEW and
builds catalog.sqlite (FTS5 + lookup indexes).
All commands run inside the CloudBeaver WSL distro. Scripts avoid PowerShell→wsl quoting.
# 1. copy source data into ./data (once)
bash scripts/copy-data.sh
# 2. install deps, build TS, compile catalog.sqlite
bash scripts/install-build-compile.sh
# 3. boot + live smoke test (starts server, runs a real MCP client, asserts acceptance criteria)
bash scripts/verify.shdocker compose build
docker compose up -d
curl http://localhost:8979/healthThe multi-stage image compiles the catalog at build time and ships only dist/ +
catalog.sqlite (the 250MB CSVs are dropped from the runtime image).
Point any MCP client at the Streamable-HTTP endpoint:
http://<host>:8979/mcpe.g. Claude Agent SDK / Codex SDK mcpServers config, or the MCP inspector.
Var | Default | Meaning |
|
| HTTP port |
|
| bind host |
|
| catalog path |
|
| source CSV/JSON dir (compile step) |
|
| smoke-client target |
This server cannot be deployed
List datasets, schemas, run APL queries, and use prompts for exploration, anomalies, and monitoring.
PostgreSQL, MySQL, OpenAPI/Swagger, and shared Agent Memory with scoped access.
Shared, permission-aware company context for AI agents, with provenance, approvals and audit.
MCP facade over the Nebelus Construction API. ~48 tools give full agent build parity: create/update/probe agents, edit graphs, attach knowledge and vector stores, wire connectors, set governance policies and locked guardrails, enable grounding-trace, and read deployment wiring. Purpose-built for regulated industries: data residency is enforced per region (EU / GCC-KSA), with PII controls and an audit trail. Agents are created as drafts — no deploy tool is exposed over MCP by design; publishing happens in the Nebelus console.
