Connect Claude Code — or any MCP-capable agent — to Volt's native MCP server. Your agent reads and writes real transactional state through eight built-in tools, so it answers with facts instead of guesses.
QUICK START
The Volt MCP server is a single JAR that any MCP client can launch over stdio. Register it once and your agent can explore the schema, query live data, and call stored procedures.
Volt + VMC + the MCP JAR
One command, once
Dedicated role, read-only first
It picks the right tool
You need a running Volt database with the Active(MCP) license feature, its VMC service (the JSON/HTTP API) reachable from wherever your agent runs, and the Volt MCP server JAR.
# You need three things: # 1. An MCP-capable AI agent, e.g. Claude Code # 2. A running Volt database with the Active(MCP) license # 3. Its VMC service (JSON/HTTP API), e.g. http://vmcserver:8080 $ ls ~/voltmcp volt-mcp-server.jar
Where do I get the JAR? The MCP server ships with Volt Enterprise — ask your Volt contact, or start here if you're new to Volt. Full details: Installing the VoltDB MCP Server.
One command adds Volt to your agent's list of MCP servers. Claude Code is shown here; any MCP client that can launch a stdio server works the same way.
$ claude mcp add voltdb -- \
java -jar ~/voltmcp/volt-mcp-server.jar \
--url http://vmcserver:8080Once is enough. The Volt tools are available in every later session. If your VMC service uses TLS, point --url at https:// and see the SSL configuration options.
Treat your agent like any other client: give it its own account and start read-only. Create a role with SQL read permissions, keep the credentials in a file, and pass them to the server.
$ sqlcmd 1> CREATE ROLE mcp WITH SQLREAD; $ cat $HOME/voltcred.txt username: mcpserver password: ******** $ claude mcp add voltdb -- \ java -jar ~/voltmcp/volt-mcp-server.jar \ --url http://vmcserver:8080 \ --credentials=$HOME/voltcred.txt
Widen later. When you trust the agent's behavior, expand the mcp role's permissions — writes and procedure calls included. Details: Configuring the MCP Server.
No SQL in your prompt. The agent discovers your schema and picks the right tool — get_tables to look around, execute_sql to answer, call_procedure to act.
> Which account had the highest spend in the last 5 minutes? ⏺ get_tables() → transactions, accounts ⏺ get_table_schema("transactions") ⏺ execute_sql("SELECT account_id, SUM(amount) …") Account 42 — $18,304.77 across 214 transactions.
That's it. The answer came from live state, not an embedding. See How to Use the VoltDB MCP Server for prompting patterns.
WHAT YOUR AGENT CAN DO
The MCP server exposes Volt to your agent as ready-made tools — schema discovery, ad-hoc SQL, stored procedures, and cluster introspection.
Run any SQL statement against live data — DDL included.
List the tables and views in the database.
Column names and datatypes for any table.
Discover the stored procedures available to call.
Invoke a stored procedure with arguments — reads or writes.
Cluster version, node status, and deployment config.
Live metrics on the cluster and its components.
What the connected account is allowed to do.
WHAT YOU CAN BUILD
Where agents need authoritative, real-time data — and the ability to act on it.
An agent that checks stock, reorders, and answers "how many, right now?" with the real number — then writes the decision back with ACID guarantees.
Keep embeddings for your documents; pull balances, quotas, and session state from Volt at answer time. No re-embedding cycle, no stale context.
Agents that approve orders, adjust limits, or route tickets — every action a serializable transaction your auditors can trace.
Agents don't just need intelligence — they need an operational layer that's correct, consistent, and current at the moment they act. Volt is the authoritative state your agents query during reasoning, so decisions run on live business truth instead of a stale snapshot.
The Model Context Protocol makes that connection standard: one server, eight tools, and your agent can read state, execute transactions, and write back decisions with ACID guarantees — a million perfect actions every second.
More Volt videos →Before your agent touches live data, teach it how Volt works. Agent Skills are free, MIT-licensed instruction sets that make coding agents fluent in Volt — partitioning, stored procedures, deployment, the lot.
Drop a skill into your agent's skills directory and it writes Volt applications the way Volt engineers do — correctly partitioned schemas, single-partition procedures, tested clients.
Get the skills on GitHub →# Install a skill (Claude Code shown) $ git clone https://github.com/VoltDB/volt-skills $ cp -r volt-skills/voltdb-development ~/.claude/skills/ # Use it $ claude > /voltdb-development > Build me a spend-velocity check for this schema.
REFERENCE
Chapter overview
Prerequisites and setup
Credentials, TLS, security roles
The eight tools and how agents use them
Installing and using skills
Free, MIT-licensed agent skills
CONNECT
Get help, share what you've built, and shape the future of Volt.
EXPLORE OTHER PATHS
Connect Volt to your Kafka cluster and make real-time decisions on streaming data in under 10ms.
Go to Kafka Developer → TransactionsMicrosecond ACID transactions with familiar SQL. Replace PostgreSQL + Redis in 5 minutes.
Go to DB Developer →Go from zero to an MCP-connected agent in minutes. Agent Skills are free, and the sandbox runs in your browser.