Assistants search and read RefLib corpora by section, not from model memory.
Maintain this server? Claim it
Call this server's tools from your browser. Nothing to install.
Bearer <your API key>, for search and read. Free key at https://www.reflib.online/register
Requests go through mcptry because browsers can't call most MCP servers directly. Header values are sent to the server for that request only and are never stored or logged.
Tool list from our last check. Connect live to refresh it.
List every corpus in the Postgres database with id, name, and description. Use the `id` from the listing as search's corpus argument.
This tool takes no arguments.
This tool isn't marked read-only, so it may change something on the server or in an account you connected.
Pick your client. Values in capitals, like YOUR_TOKEN, are yours to fill in.
Run in a terminal
claude mcp add --transport http reflib https://mcp.reflib.online/mcp
Headers
Authorization (optional, not in the config above): Bearer <your API key>, for search and read. Free key at https://www.reflib.online/registerList every corpus in the Postgres database with id, name, and description. Use the `id` from the listing as search's corpus argument.
No arguments.
{
"type": "object",
"properties": {},
"additionalProperties": false
}Read a section by ID ({corpus}:{hierarchy}).
| Argument | Type | Description |
|---|---|---|
| section_id* | string | e.g. postgresql_16:4.5.3.1 |
{
"type": "object",
"required": [
"section_id"
],
"properties": {
"section_id": {
"type": "string",
"description": "e.g. postgresql_16:4.5.3.1"
}
},
"additionalProperties": false
}Search a RefLib corpus. Then read section IDs you need.
| Argument | Type | Description |
|---|---|---|
| corpus* | string | Corpus id from list_corpora |
| question* | string | Precise question |
| n_results | integer | Maximum results (default 5) |
{
"type": "object",
"required": [
"corpus",
"question"
],
"properties": {
"corpus": {
"type": "string",
"description": "Corpus id from list_corpora"
},
"question": {
"type": "string",
"description": "Precise question"
},
"n_results": {
"type": "integer",
"description": "Maximum results (default 5)"
}
},
"additionalProperties": false
}We read every tool's name, description and inputs the way the model will, looking for instructions hidden from you. A clean result means no warning signs, not a guarantee. Scanned Nov 16, 2026.
No readOnlyHint / destructiveHint annotations, so clients can't tell a read from a write without asking.
Fix: Add annotations to each tool, e.g. { readOnlyHint: true } for tools that only read.