Skip to main content




You can use the dbt-rpc plugin to run a Remote Procedure Call (rpc) dbt server. This server compiles and runs queries in the context of a dbt project. Additionally, the RPC server provides methods that enable you to list and terminate running processes. We recommend running an rpc server from a directory containing a dbt project. The server will compile the project into memory, then accept requests to operate against that project's dbt context.


The dbt-rpc plugin will be fully deprecated by the end of 2022.

For now, dbt Labs actively maintains and uses dbt-rpc to enable interactive dbt development. Once we announce the next-generation dbt Server is available for general release, we will deprecate the legacy plugin and only fix critical issues for a period of six months. After six months, we will archive this repository for read-only use.

Running on Windows

We do not recommend running the rpc server on Windows because of reliability issues. A Docker container may provide a useful workaround, if required.

For more details, see the dbt-rpc repository source code.

Running the server:

$ dbt-rpc serveRunning with dbt=0.15.0
16:34:31 | Concurrency: 8 threads (target='dev')16:34:31 |16:34:31 | Done.Serving RPC server at requests to http://localhost:8580/jsonrpc

Configuring the server

  • --host: Specify the host to listen on (default=
  • --port: Specify the port to listen on (default=8580)

Submitting queries to the server: The rpc server expects requests in the following format:

{    "jsonrpc": "2.0",    "method": "{ a valid rpc server command }",    "id": "{ a unique identifier for this query }",    "params": {        "timeout": { timeout for the query in seconds, optional },    }}

Built-in Methods


The status method will return the status of the rpc server. This method response includes a high-level status, like ready, compiling, or error, as well as the set of logs accumulated during the initial compilation of the project. When the rpc server is in the compiling or error state, only built-in methods of the RPC server will be accepted.

Example request

{    "jsonrpc": "2.0",    "method": "status",    "id": "2db9a2fe-9a39-41ef-828c-25e04dd6b07d"}

Example response

{    "result": {        "status": "ready",        "error": null,        "logs": [..],        "timestamp": "2019-10-07T16:30:09.875534Z",        "pid": 76715    },    "id": "2db9a2fe-9a39-41ef-828c-25e04dd6b07d",    "jsonrpc": "2.0"}


The poll endpoint will return the status, logs, and results (if available) for a running or completed task. The poll method requires a request_token parameter which indicates the task to poll a response for. The request_token is returned in the response of dbt tasks like compile, run and test.


  • request_token: The token to poll responses for
  • logs: A boolean flag indicating if logs should be returned in the response (default=false)
  • logs_start: The zero-indexed log line to fetch logs from (default=0)

Example request

{    "jsonrpc": "2.0",    "method": "poll",    "id": "2db9a2fe-9a39-41ef-828c-25e04dd6b07d",    "params": {        "request_token": "f86926fa-6535-4891-8d24-2cfc65d2a347",        "logs": true,        "logs_start": 0    }}

Example Response

{    "result": {        "results": [],        "generated_at": "2019-10-11T18:25:22.477203Z",        "elapsed_time": 0.8381369113922119,        "logs": [],        "tags": {            "command": "run --select my_model",            "branch": "abc123"        },        "status": "success"    },    "id": "2db9a2fe-9a39-41ef-828c-25e04dd6b07d",    "jsonrpc": "2.0"}


The ps methods lists running and completed processes executed by the RPC server.


  • completed: If true, also return completed tasks (default=false)

Example request:

{    "jsonrpc": "2.0",    "method": "ps",    "id": "2db9a2fe-9a39-41ef-828c-25e04dd6b07d",    "params": {        "completed": true    }}

Example response:

{    "result": {        "rows": [            {                "task_id": "561d4a02-18a9-40d1-9f01-cd875c3ec56d",                "request_id": "3db9a2fe-9a39-41ef-828c-25e04dd6b07d",                "request_source": "",                "method": "run",                "state": "success",                "start": "2019-10-07T17:09:49.865976Z",                "end": null,                "elapsed": 1.107261,                "timeout": null,                "tags": {                    "command": "run --select my_model",                    "branch": "feature/add-models"                }            }        ]    },    "id": "2db9a2fe-9a39-41ef-828c-25e04dd6b07d",    "jsonrpc": "2.0"}


The kill method will terminate a running task. You can find a task_id for a running task either in the original response which invoked that task, or in the results of the ps method.

Example request

{    "jsonrpc": "2.0",    "method": "kill",    "id": "2db9a2fe-9a39-41ef-828c-25e04dd6b07d",    "params": {        "task_id": "{ the task id to terminate }"    }}

Running dbt projects

The following methods make it possible to run dbt projects via the RPC server.

Common parameters

All RPC requests accept the following parameters in addition to the parameters listed:

  • timeout: The max amount of time to wait before cancelling the request.
  • task_tags: Arbitrary key/value pairs to attach to this task. These tags will be returned in the output of the poll and ps methods (optional).

Running a task with CLI syntax


  • cli: A dbt command (eg. run --select abc+ --exclude +def) to run (required)
{    "jsonrpc": "2.0",    "method": "cli_args",    "id": "<request id>",    "params": {        "cli": "run --select abc+ --exclude +def",        "task_tags": {            "branch": "feature/my-branch",            "commit": "c0ff33b01"        }    }}

Several of the following request types accept these additional parameters:

  • threads: The number of threads to use when compiling (optional)
  • select: The space-delimited set of resources to execute (optional). (models is also supported on some request types for backwards compatibility.)
  • selector: The name of a predefined YAML selector that defines the set of resources to execute (optional)
  • exclude: The space-delimited set of resources to exclude from compiling, running, testing, seeding, or snapshotting (optional)
  • state: The filepath of artifacts to use when establishing state (optional)

Compile a project (docs)

{    "jsonrpc": "2.0",    "method": "compile",    "id": "<request id>",    "params": {            "threads": "<int> (optional)",            "select": "<str> (optional)",            "exclude": "<str> (optional)",            "selector": "<str> (optional)",            "state": "<str> (optional)"        }}

Run models (docs)

Additional parameters:

  • defer: Whether to defer references to upstream, unselected resources (optional, requires state)
{    "jsonrpc": "2.0",    "method": "run",    "id": "<request id>",    "params": {            "threads": "<int> (optional)",            "select": "<str> (optional)",            "exclude": "<str> (optional)",            "selector": "<str> (optional)",            "state": "<str> (optional)",            "defer": "<bool> (optional)"        }}

Run tests (docs)

Additional parameters:

  • data: If True, run data tests (optional, default=true)
  • schema: If True, run schema tests (optional, default=true)
{    "jsonrpc": "2.0",    "method": "test",    "id": "<request id>",    "params": {            "threads": "<int> (optional)",            "select": "<str> (optional)",            "exclude": "<str> (optional)",            "selector": "<str> (optional)",            "state": "<str> (optional)",            "data": "<bool> (optional)",            "schema": "<bool> (optional)"        }}

Run seeds (docs)


  • show: If True, show a sample of the seeded data in the response (optional, default=false)
{    "jsonrpc": "2.0",    "method": "seed",    "id": "<request id>",    "params": {            "threads": "<int> (optional)",            "select": "<str> (optional)",            "exclude": "<str> (optional)",            "selector": "<str> (optional)",            "show": "<bool> (optional)",            "state": "<str> (optional)"        }}

Run snapshots (docs)

{    "jsonrpc": "2.0",    "method": "snapshot",    "id": "<request id>",    "params": {            "threads": "<int> (optional)",            "select": "<str> (optional)",            "exclude": "<str> (optional)",            "selector": "<str> (optional)",            "state": "<str> (optional)"        }}

Build (docs)

{    "jsonrpc": "2.0",    "method": "build",    "id": "<request id>",    "params": {            "threads": "<int> (optional)",            "select": "<str> (optional)",            "exclude": "<str> (optional)",            "selector": "<str> (optional)",            "state": "<str> (optional)",            "defer": "<str> (optional)"        }}

List project resources (docs)

Additional parameters:

  • resource_types: Filter selected resources by type

  • output_keys: Specify which node properties to include in output

    {    "jsonrpc": "2.0",    "method": "build",    "id": "<request id>",    "params": {        "select": "<str> (optional)",        "exclude": "<str> (optional)",        "selector": "<str> (optional)",        "resource_types": ["<list> (optional)"],        "output_keys": ["<list> (optional)"],    }}

Generate docs (docs)

Additional parameters:

  • compile: If True, compile the project before generating a catalog (optional, default=false)
{    "jsonrpc": "2.0",    "method": "docs.generate",    "id": "<request id>",    "params": {            "compile": "<bool> (optional)",            "state": "<str> (optional)"        }}

Compiling and running SQL statements

Compiling a query

This query compiles the sql select {{ 1 + 1 }} as id (base64-encoded) against the rpc server:

{    "jsonrpc": "2.0",    "method": "compile_sql",    "id": "2db9a2fe-9a39-41ef-828c-25e04dd6b07d",    "params": {        "timeout": 60,        "sql": "c2VsZWN0IHt7IDEgKyAxIH19IGFzIGlk",        "name": "my_first_query"    }}

The resulting response will include a key called compiled_sql with a value of 'select 2'.

Executing a query

This query executes the sql select {{ 1 + 1 }} as id (bas64-encoded) against the rpc server:

{    "jsonrpc": "2.0",    "method": "run_sql",    "id": "2db9a2fe-9a39-41ef-828c-25e04dd6b07d",    "params": {        "timeout": 60,        "sql": "c2VsZWN0IHt7IDEgKyAxIH19IGFzIGlk",        "name": "my_first_query"    }}

The resulting response will include a key called table with a value of {'column_names': ['?column?'], 'rows': [[2.0]]}

Reloading the Server

When the dbt Server starts, it will load the dbt project into memory using the files present on disk at startup. If the files in the dbt project should change (either during development or in a deployment), the dbt Server can be updated live without cycling the server process. To reload the files present on disk, send a "hangup" signal to the running server process using the Process ID (pid) of the running process.

Finding the server PID

To find the server PID, either fetch the value from the status method response on the server, or use ps:

# Find the server PID using `ps`:ps aux | grep 'dbt-rpc serve' | grep -v grep

After finding the PID for the process (eg. 12345), send a signal to the running server using the kill command:

kill -HUP 12345

When the server receives the HUP (hangup) signal, it will re-parse the files on disk and use the updated project code when handling subsequent requests.