diff --git a/wolfram/Dockerfile b/wolfram/Dockerfile
index 0eececd..565ce73 100644
--- a/wolfram/Dockerfile
+++ b/wolfram/Dockerfile
@@ -21,6 +21,7 @@ RUN case $(uname -m) in \
 FROM --platform=linux/amd64 ${WWE}
 
 ARG WOLFRAM_ENGINE_VERSION=15.0
+ARG TOOLKIT_WOLFRAM_VERSION=v1.1.0
 
 ENV LOG_FORMAT="production"
 
@@ -30,6 +31,24 @@ WORKDIR /app
 # add wstp server to path
 ENV PATH=$PATH:/usr/local/Wolfram/WolframEngine/${WOLFRAM_ENGINE_VERSION}/SystemFiles/Links/WSTPServer
 
+# add the shared Wolfram evaluation-function toolkit (JSON comms layer),
+# used by all Wolfram-based evaluation functions via PacletDirectoryLoad.
+RUN apt-get update && apt-get install -y --no-install-recommends git ca-certificates \
+    && rm -rf /var/lib/apt/lists/*
+RUN git clone --branch ${TOOLKIT_WOLFRAM_VERSION} --depth 1 \
+      https://github.com/lambda-feedback/toolkit-wolfram.git /opt/lambda-feedback/toolkit-wolfram
+
+ENV LF_TOOLKIT_PATH="/opt/lambda-feedback/toolkit-wolfram"
+
+# Every Wolfram evaluation function follows the same shape: an evaluate.m
+# and preview.m providing evaluate`EvaluationFunction / preview`PreviewFunction.
+# Bootstrap.wl (shipped by toolkit-wolfram, cloned above) is the one fixed
+# entry point that wires those into the toolkit's ServeEvaluationFunction --
+# evaluation function repos only need to COPY their evaluate.m/preview.m
+# into WORKDIR, not their own entry script.
+ENV FUNCTION_COMMAND="wolframscript"
+ENV FUNCTION_ARGS="-f,/opt/lambda-feedback/toolkit-wolfram/Bootstrap.wl"
+
 # add shimmy
 COPY --from=shimmy /shimmy /usr/local/bin/shimmy
 
diff --git a/wolfram/README.md b/wolfram/README.md
index 4efb46e..d3537f7 100644
--- a/wolfram/README.md
+++ b/wolfram/README.md
@@ -4,14 +4,23 @@ This is the base image which is intended to be used by evaluation functions writ
 
 ## Usage
 
-This image is not meant to be run directly. Instead, it ought to be used as a base image for evaluation functions written in Wolfram Language. The evaluation function should be placed in the `/app` directory.
+This image is not meant to be run directly. Instead, it ought to be used as a base image for evaluation functions written in Wolfram Language.
 
-Here is an example of a `Dockerfile` that uses this image as a base:
+Every Wolfram evaluation function follows the same shape: an `evaluate.m` and
+`preview.m` defining `evaluate\`EvaluationFunction` and
+`preview\`PreviewFunction` respectively. This image already wires those into
+the shared [`toolkit-wolfram`](https://github.com/lambda-feedback/toolkit-wolfram)
+comms layer via `FUNCTION_COMMAND`/`FUNCTION_ARGS` (pointing at
+`toolkit-wolfram`'s `Bootstrap.wl`), so a consuming evaluation function repo
+only needs to provide those two files:
 
 ```Dockerfile
 FROM ghcr.io/lambda-feedback/evaluation-function-base/wolfram:latest
 
-COPY ./function.wl /app/function.wl
-
-CMD ["shimmy", "-i", "file", "-c", "wolframscript", "-a", "-f", "-a", "/app/function.wl"]
+COPY ./evaluate.m /app/evaluate.m
+COPY ./preview.m /app/preview.m
 ```
+
+A repo that needs custom wiring instead of the standard `evaluate.m`/
+`preview.m` convention can still override `FUNCTION_COMMAND`/`FUNCTION_ARGS`
+to point at its own entry script.