GhostCrab MCP + mindBrain SQLite — structured domain navigation for Agno agents

[FrancoisLamotte]

What problem this solves

Agno agents are powerful in isolation. The problem starts the moment two agents share a domain.

Each Agent wires its own memory backend: session storage, user memories, and summaries
all live in separate tables, scoped per agent. On a multi-agent team, no agent sees
what the others know. Facts extracted by ResearchAgent are invisible to WriterAgent.
Context rebuilt from scratch at every run. Knowledge scattered and ephemeral.

This request adds two complementary components that solve this without touching Agno core.

---

What a domain looks like in practice

A domain is any bounded context where agents need to
reason over structured relationships : not just retrieve text.

mindBrain separates two levels:

Ontology (the model) : the schema of a domain:
entity types, relationship types, constraints, vocabularies.
Defined once, shared across every agent in the team, across every run.

Knowledge Graph (the projected instance) : the populated graph:
real entities, real edges, real state : queryable via projections.

Example: multi-agent project delivery

Ontology (model)              Knowledge Graph (instance)
─────────────────             ──────────────────────────
Project                       "Platform Modernization Q3"
└── has Epic                  └── Epic: "Auth Refactor"
      └── has Task                  └── Task: "Rotate JWT keys"
            └── assigned_to               assigned_to: DevAgent
            └── status                    status: blocked
            └── depends_on          Task: "Define ERP contract"
Component                     Component: "Auth Service"
└── constrained_by            ADR: "Stateless JWT over Redis"
Agent                         OrchestratorAgent
└── owns Task                       → owns "Rotate JWT keys"
└── produces Finding          Finding: "JWT rotation requires ERP scope"

OrchestratorAgent queries MindBrain:
"what is blocking DevAgent this sprint?" →
returns the full dependency chain in a single call.
No context-window dump. No re-reading artifact files.

Example: ERP domain

Ontology (model)              Knowledge Graph (instance)
─────────────────             ──────────────────────────
Account                       "Acme Corp" (enterprise)
└── has CostCenter            └── CostCenter: "Engineering"
└── has ApprovalFlow          └── ApprovalFlow: PO > €5k → CFO
GLEntry                       GL-20260501: amount=€12,400
└── linked_to CostCenter            linked_to: "Engineering"

A FinanceAgent resolving a billing anomaly navigates
account structure, approval rules, and GL history
without re-injecting 300 rows of CSV each session.

Example: CRM domain

Ontology (model)              Knowledge Graph (instance)
─────────────────             ──────────────────────────
Account                       "NovaTech" (tier=enterprise)
└── has Contact               └── Contact: "Maria Chen" (VP Engineering)
└── has Deal                  └── Deal: "Platform License" (stage=negotiation)
└── has Interaction           └── Call: 2026-05-14, sentiment=positive

A SalesAgent preparing a renewal call retrieves
account tier, active deals, recent interaction history,
and known constraints in one structured query.

Example: Customer Support domain

Ontology (model)              Knowledge Graph (instance)
─────────────────             ──────────────────────────
Ticket                        TK-9821: priority=critical
└── linked_to Component       └── linked_to: "Auth Service"
└── references Incident       └── Incident: "JWT outage 2026-04-30"
└── assigned_to Agent         └── SupportAgent-3
KnownIssue                    KI-44: "Token expiry on clock skew"
└── affects Component               affects: "Auth Service"

A SupportAgent resolving TK-9821 immediately sees
the linked incident, the known issue, and its resolution history —
without searching knowledge bases or escalating blindly.

---

The meta-ontology: where mindBrain becomes strategic

Each domain above is useful alone. The real leverage is connecting them.

         ERP ←──────────────────┐
                                 │
         CRM ←─────── mindBrain ─┼──→ Project Management
                                 │
         Support ←───────────────┘
                                 │
                                 └──→ Knowledge / Documentation

A single Agno team working on a cross-domain task — say, an enterprise renewal
involving a billing dispute, an open support ticket, and a stalled deal —
can query all four domains in one pass from the same structured registry.

The Agno OrchestratorAgent doesn't ask each specialized agent
"what do you know about NovaTech?" — it navigates the graph directly.
Context does not need to transit through natural language.
Relations carry their semantics natively.

That is the difference between memory and ontology.

---

Two components, one integration

mindBrain (SQLITE : Personal edition)

The data layer. It organizes domain knowledge into three constructs:

• Facets : typed dimensions that structure a domain (status, priority, tier, domain, agent_id, sprint, ...)
• Semantic graphs : edges that capture typed relationships between entities (ASSIGNED_TO, DEPENDS_ON, REFERENCES, CONSTRAINS, KNOWS_ABOUT, ...)
• Projections : pre-computed views of complex queries, surfaced instantly at zero inference cost

Agents don't search through data. They navigate a structured domain
the way a developer navigates a codebase, not a search index.
Embedding search is available for unstructured recall, but it's the fallback,
not the default.

GhostCrab MCP

The gateway layer. An MCP server that gives Agno agents the tools to:

• Read and write the shared ontology registry without direct database access
• Execute faceted search, graph traversal, and projection queries
• Push memory events from Agno runs into the durable knowledge graph

GhostCrab is a protocol bridge. mindBrain exists and operates independently.

Agno Agent / Team
└── GhostCrab MCP  (streamable-http or stdio)
      └── mindBrain SQLITE
            ├── facets       — typed domain dimensions
            ├── graphs       — semantic edges across domains
            └── projections  — shared context for any agent at any time

---

Why this belongs outside Agno core

It should stay external.

Agno already exposes clean extension points: MCPTools for tool-level integration
and a pluggable MemoryDb interface for deep memory integration.
mindBrain uses both without modifying Agno agents, teams, or workflows.

Domain ontologies are specific to each application's context.
Baking a particular schema into the framework would be the wrong abstraction.

GhostCrab's value is precisely that it's optional, composable,
and invisible when not needed.

If GhostCrab is absent : Agno behaves exactly as before.
If mindBrain is empty : agents start with a blank namespace and populate it
through normal tool use during the first run.

---

Setup

Two integration paths, independently usable and combinable:

Path 1 — MCPTools (preferred, zero friction)

from agno.agent import Agent
from agno.models.anthropic import Claude
from agno.tools.mcp import MCPTools

agent = Agent(
    name="ResearchAgent",
    model=Claude(id="claude-sonnet-4-5"),
    tools=[
        MCPTools(
            transport="streamable-http",
            url="http://localhost:8080/mcp"  # GhostCrab MCP endpoint
        )
    ],
    enable_user_memories=False,  # MindBrain is the single registry
)

The agent gets tools such as ontology_query, entity_resolve, graph_traverse,
context_push from GhostCrab — without changing the framework.

Path 2 — Custom MemoryDb backend (deep integration)

from agno.memory.v2.memory import Memory
from ghostcrab_agno import MindBrainMemoryDb  # thin wrapper package

memory = Memory(
    model=Claude(id="claude-haiku-3-5"),
    db=MindBrainMemoryDb(
        mindbrain_dsn="sqlite://localhost/mindbrain",
        namespace="team_alpha"
    )
)

agent = Agent(
    model=Claude(id="claude-sonnet-4-5"),
    memory=memory,
    enable_user_memories=True,  # writes straight into MindBrain after each run
)

Agno fact extraction writes directly into the ontology registry —
no explicit tool call needed from the agent.

→ Full configuration, skill files, and tested Agno walkthrough:
https://github.com/mindflight-orchestrator/ghostcrab-personal-mcp/blob/main/ghostcrab-integrations/agno/sop-agno-ghostcrab-mindbrain.md

https://github.com/mindflight-orchestrator/ghostcrab-personal-mcp/blob/main/ghostcrab-integrations/agno/skill_ghostcrab_personal_agno_runtime.md

---

Scope of this request

• Adds GhostCrab MCP + mindBrain SQLITE (Personal edition) to the Agno integrations catalog
• No changes to Agno core agents, teams, memory interfaces, or workflows
• No new required dependencies — MCPTools integration requires only agno>=0.7 (already available)
• No required migration path — existing Agno projects add GhostCrab as one additional tool

Repository	https://github.com/mindflight-orchestrator/ghostcrab-personal-mcp
License	Apache 2.0
Transport	stdio, cli
Backend	SQLITE for personal (bundled via mindBrain Personal) and PostgreSQL for the pro licence
Agno compatibility	agno>=0.7, MCPTools + MemoryDb interface
Formerly known as	Phidata integration (same API surface)