MCP Tools Reference

The Fusemomo MCP Server exposes 5 tools that your agent can call directly. All tools communicate with https://api.fusemomo.com using your API key.

Recommended Workflow

1. resolve_entity         → get canonical entity_id
2. get_recommendation     → find the best action (Builder+)
3. [Agent performs action]
4. log_interaction        → record the outcome
5. update_recommendation_outcome → close the feedback loop

---

resolve_entity

Always call this first. Maps any known identifiers to a canonical entity, creating it if none exists. Returns the entity_id needed by all other tools.

Parameters

Name	Required	Description
identifiers	✅	Object: { "source_key": "identifier_value" }
display_name	—	Human-readable label for this entity
entity_type	—	Category (e.g. user, ticket, account)
metadata	—	Any additional JSON data

Example

"Resolve the entity for client_id=ibergx9H50 and email=[EMAIL_ADDR]"
→ { "entity_id": "ent_abc123", "behavioral_score": 0.72, "total_interactions": 14, ... }

---

get_recommendation

Returns the highest-success action for an entity, based on historical outcome data. Call this before deciding what your agent should do.

Requires Builder plan.

Parameters

Name	Required	Description
entity_id	✅	Canonical entity UUID
intent	✅	Goal scope (e.g. code_review, support_escalation)
lookback_days	—	History window, default 90

Example

"What is the best action for entity ent_abc123 with intent 'code_review'?"
→ { "data_sufficient": true, "confidence_score": 0.87, "primary": { "api": "github", "action_type": "assign_reviewer", ... } }

---

log_interaction

Records a behavioral event in the graph. Call this after any action your agent takes through an external API.

Parameters

Name	Required	Description
entity_id	✅	Canonical entity UUID
agent_id	✅	Your agent's identifier
api	✅	External service used
action_type	✅	Normalized action category
action	✅	Specific description of the action
outcome	✅	success | failed | pending | ignored | unknown
intent	✅	Goal of the action
metadata	—	Additional JSON

Example

"Log that agent_review_bot assigned PR #42 to octocat via github, outcome: success"
→ { "interaction_id": "int_xyz", "logged_at": "..." }

---

update_recommendation_outcome

Closes the ML feedback loop after a recommendation was acted on (or ignored). Must be called with the recommendation_id returned by get_recommendation.

Parameters

Name	Required	Description
recommendation_id	✅	UUID from get_recommendation
was_followed	✅	Did the agent act on the primary recommendation?
outcome_interaction_id	—	Link to the logged interaction for this action

Example

"Mark recommendation rec_456 as followed, linked to interaction int_789"
→ { "recommendation_id": "rec_456", "was_followed": true, "updated_at": "..." }

---

get_entity

Fetches the complete profile of a resolved entity — all linked identifiers, interaction history summary, and behavioral score.

Parameters

Name	Required	Description
entity_id	✅	Canonical entity UUID

Example

"Get full profile for entity ent_abc123"
→ { "id": "ent_abc123", "behavioral_score": 0.72, "identifiers": [...], ... }