Data Tools
Browse and query your project's databases, and fire off API requests — all from your phone, executing on your paired desktop. AI agents get scoped access to the same connections through read-gated MCP tools.
Database Browser
Add a database connection once in the desktop app, then browse it from your phone over the same encrypted relay connection as everything else in Hermes.
Schema Inspection
See tables, columns, types, and constraints for any connected database.
Data Inspection
Browse table rows with pagination — no query required to look around.
Per-Project Connections
Each project keeps its own connection list, so switching projects switches your databases too.
Stored on Desktop
Connections and encrypted credentials live in the desktop app's local SQLite store. Mobile browses the same data over the relay — it never stores the raw credentials.
What you need
Database connection details — host, port, and credentials — added once in the desktop app.
Database Querying
Go beyond browsing: run a real query against a connected database and get the results back as a table, right on your phone.
Execute Queries
Run a query against any connected database straight from the browser — no separate SQL client required.
MySQL and Elasticsearch
The query panel runs MySQL statements and Elasticsearch searches against your connections.
Tabular Results
Rows and columns render in a scrollable table, with row counts and timing for every run.
Query History
Every query you run is kept per connection, alongside named queries you save and favorite for later.
Database MCP Access
AI agents get their own scoped door into the same connections through the
hermes_db MCP server — no separate setup, no copy-pasted credentials.
- Ask in plain language — the agent writes and runs the query for you.
- MySQL queries run inside a
START TRANSACTION READ ONLYguard; writes are rejected unless the connection has AI Write Mode turned on. - Elasticsearch access covers index search and mapping inspection.
- Each AI CLI only sees the database connections that belong to the current project.
What you need
A database connection already configured, plus a subscription for whichever AI CLI you use — Claude Code, Codex, or Gemini CLI. Runs on the paired desktop.
hermes_db is one of Hermes' built-in MCP servers. See the full list on the
Toolbox page.
API Requests Client
A lightweight REST client built into Hermes — save a request once, then run it from wherever you are.
Saved Collections
Organize requests into per-project collections, with folders for grouping related endpoints.
Variable Resolution
Set variables per environment and reference them in any request — Hermes resolves them before it goes out.
Full Response View
Inspect status code, headers, timing, and body — pretty-printed or raw — for every response.
What you need
The API's endpoint URLs and any credentials the requests need — API keys, bearer tokens, or OAuth.
Requests MCP Access
The hermes_requests MCP hands AI agents the same saved collections, without
ever handing over the underlying auth.
- List collections and requests, or execute a saved request with variables resolved from the active environment.
- Auth values and variables are never returned to the agent — only the resolved request and its response are.
- Mutating methods (anything beyond GET, HEAD, or OPTIONS) stay blocked per collection until you turn on "AI can execute mutating requests".
What you need
A collection with at least one saved request, plus a subscription for whichever AI CLI you use — Claude Code, Codex, or Gemini CLI. Runs on the paired desktop.
hermes_requests is one of Hermes' built-in MCP servers. See the full list on the
Toolbox page.
Supported Engines
Data Tools targets two engines end-to-end — connecting, browsing, querying, and, for AI agents, MCP access all work the same way.
- MySQL — queries, schema browsing (tables, columns, indexes, foreign keys), and row-level data browsing.
- Elasticsearch — index listing, mapping inspection, and search.
Hermes also uses SQLite internally, on desktop, to store your connection metadata. That's an implementation detail, not a database engine you connect to or query.