Knowledge MCP
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| KNOWLEDGE_MCP_INDEX | No | The index backend to use: 'graphiti' (semantic retrieval, default) or 'memory' (lexical, prefix matching). | graphiti |
| KNOWLEDGE_MCP_PROJECT | No | A fixed project root path, overriding the working directory discovery. | |
| KNOWLEDGE_MCP_NEO4J_PASSWORD | No | Password for the Neo4j graph database used by the graphiti backend. |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": false
} |
| prompts | {
"listChanged": false
} |
| resources | {
"subscribe": false,
"listChanged": false
} |
| experimental | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| start_taskA | Chame ANTES de comecar qualquer tarefa neste projeto. Devolve o conhecimento acumulado que e relevante para ela: regras do projeto, decisoes ja tomadas, preferencias e convencoes. Somente leitura. |
| contextB | Consulta livre ao conhecimento do projeto, para quando a pergunta nao e o inicio de uma tarefa. Ex.: 'o que sabemos sobre autenticacao?'. Somente leitura. |
| finish_taskA | Chame ao TERMINAR uma tarefa, passando um resumo do que foi feito e decidido. Analisa o resumo e SUGERE o que merece virar conhecimento permanente. NAO grava nada. Apresente o resultado ao usuario e, somente se ele aprovar, chame |
| rememberA | Grava conhecimento na base do projeto. Use com |
| searchB | Procura no conhecimento registrado do projeto (decisoes, entidades, preferencias, convencoes, tecnologias). Somente leitura. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 5 tools
The tools are mostly distinct: search/context are read-only queries (with context being more natural-language free-form), start_task/finish_task are lifecycle hooks, and remember is the only write operation. The main ambiguity is between search and context, which both retrieve project knowledge, though descriptions differentiate them reasonably.
Tool names follow a clean snake_case verb pattern (search, context, remember, start_task, finish_task). However, they mix verbs and nouns inconsistently - two are bare verbs (context, remember, search) while two are verb_noun compound tasks (start_task, finish_task). The naming is readable but not patterned.
Five tools is ideal for a knowledge-management MCP server. Each tool maps to a clear function: querying (search/context), task lifecycle (start/finish), and writing (remember). There's no bloat or redundancy in the count.
The surface covers the full knowledge-management lifecycle: read (search/context), write (remember), and contextual hooks (start/finish_task). A minor gap is the absence of an explicit delete/update operation for knowledge records - forget or edit are missing - but the core workflow is well covered.