Usage — API / CLI / MCP

egeria-mcp exposes the same capability three ways: as MCP tools an agent calls, as a Python API (EgeriaApi) you import, and as a harvest CLI. The complete tool surface and the verified Egeria 6.0 REST contract are in Overview.

As an MCP server

Once deployed, the server registers 21 tools. Reads work with no configuration beyond the platform connection; writes require EGERIA_ENABLE_WRITE=true.

Group	Tools
Granular reads	egeria_asset_search, egeria_glossary_lookup, egeria_glossary_categories, egeria_lineage, egeria_governance_for, egeria_list_policies
Broad action-dispatch reads	egeria_catalog, egeria_data_design, egeria_collection, egeria_solution, egeria_governance_catalog, egeria_actors, egeria_metadata
Routing	egeria_governed_route
Write-gated	egeria_classify, egeria_create_term, egeria_assert_lineage, egeria_harvest

Example agent prompts that map onto these tools:

• "Search the catalog for assets named like 'customer'" → egeria_asset_search
• "What is the downstream lineage impact of asset <guid>?" → egeria_lineage
• "Is it safe to act on asset <guid>?" → egeria_governed_route

As a Python API

EgeriaApi is a tolerant raw-httpx facade — every call degrades to [] / a clear error rather than raising, so it is safe to use without a reachable platform.

from egeria_mcp.api.api_client_egeria import EgeriaApi

api = EgeriaApi(
    platform_url="https://your-egeria:9443",
    view_server="qs-view-server",
    user_id="erinoverview",
    user_pwd="secret",
    verify_ssl=False,
)

# Reads
assets = api.asset_search("customer")          # flat records with GUIDs
terms = api.list_glossary_terms()
flows = api.list_data_flows()                  # DataFlow lineage edges
gov = api.governance_for(assets[0]["guid"])    # confidentiality, etc.

# The federation decision
from egeria_mcp.governed_routing import governed_route
decision = governed_route(api, assets[0]["guid"])
print(decision["decision"])                    # proceed | review | require_approval

Build a client straight from the environment:

from egeria_mcp.auth import get_client
api = get_client()        # reads EGERIA_* from the environment / .env

Writes

Writes are gated behind enable_write (env: EGERIA_ENABLE_WRITE=true):

api = EgeriaApi(enable_write=True, ...)
api.create_term("Customer", "A party that purchases goods or services")
api.classify(guid, "Confidentiality", {"confidentialityLevel": 3})
api.assert_lineage(source_guid, target_guid, label="DataFlow")

As a harvest CLI

The bottom-up harvest populates Egeria from your data estate (needs writes enabled). Run a single layer or all:

EGERIA_PLATFORM_URL=https://your-egeria:9443 EGERIA_ENABLE_WRITE=true \
  python -m egeria_mcp.harvest datastores      # the anchor layer

# Cross-link the independently harvested layers into one graph
EGERIA_ENABLE_WRITE=true python -m egeria_mcp.harvest reconcile

# Coverage report (read-only): islands + per-layer lineage %
python -m egeria_mcp.harvest audit

Each connector reads its own credentials (see .env.example) and remains inactive when those credentials are absent. The full source → Egeria-type → confidentiality map for all 34 layers is in Harvesters.