diff --git a/pyproject.toml b/pyproject.toml
index c5b48619..ad43981b 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -22,7 +22,6 @@ requires-python = ">=3.10"
 plotting = ["scipy", "matplotlib"]
 # REST service support
 service = ["fastapi>=0.100.0", "uvicorn"]
-# For development tests/docs
 dev = [
     # This syntax is supported since pip 21.2
     # https://github.com/pypa/pip/issues/10393
@@ -41,11 +40,13 @@ dev = [
     "sphinx-copybutton",
     "sphinx-design",
     "sphinxcontrib-openapi",
+    "strawberry-graphql[debug-server]",
+    "strawberry-graphql[fastapi]",
     "tox-direct",
     "types-mock",
     "httpx",
     "myst-parser",
-]
+] # For development tests/docs
 
 [project.scripts]
 scanspec = "scanspec.cli:cli"
diff --git a/src/scanspec/schema/__init__.py b/src/scanspec/schema/__init__.py
new file mode 100644
index 00000000..e69de29b
diff --git a/src/scanspec/schema/resolvers.py b/src/scanspec/schema/resolvers.py
new file mode 100644
index 00000000..999cd822
--- /dev/null
+++ b/src/scanspec/schema/resolvers.py
@@ -0,0 +1,46 @@
+from typing import Any
+
+import numpy as np
+from specs import Line
+
+from scanspec.core import (
+    Frames,
+    Path,
+)
+
+
+def reduce_frames(stack: list[Frames[str]], max_frames: int) -> Path:
+    """Removes frames from a spec so len(path) < max_frames.
+
+    Args:
+        stack: A stack of Frames created by a spec
+        max_frames: The maximum number of frames the user wishes to be returned
+    """
+    # Calculate the total number of frames
+    num_frames = 1
+    for frames in stack:
+        num_frames *= len(frames)
+
+    # Need each dim to be this much smaller
+    ratio = 1 / np.power(max_frames / num_frames, 1 / len(stack))
+
+    sub_frames = [sub_sample(f, ratio) for f in stack]
+    return Path(sub_frames)
+
+
+def sub_sample(frames: Frames[str], ratio: float) -> Frames:
+    """Provides a sub-sample Frames object whilst preserving its core structure.
+
+    Args:
+        frames: the Frames object to be reduced
+        ratio: the reduction ratio of the dimension
+    """
+    num_indexes = int(len(frames) / ratio)
+    indexes = np.linspace(0, len(frames) - 1, num_indexes, dtype=np.int32)
+    return frames.extract(indexes, calculate_gap=False)
+
+
+def validate_spec(spec: Line) -> Any:
+    """A query used to confirm whether or not the Spec will produce a viable scan."""
+    # TODO apischema will do all the validation for us
+    return spec.serialize()
diff --git a/src/scanspec/schema/schema.py b/src/scanspec/schema/schema.py
new file mode 100644
index 00000000..4f58b172
--- /dev/null
+++ b/src/scanspec/schema/schema.py
@@ -0,0 +1,107 @@
+from typing import Any
+
+import strawberry
+from fastapi import FastAPI
+from resolvers import reduce_frames, validate_spec
+from specs import PointsResponse
+from strawberry.fastapi import GraphQLRouter
+
+from scanspec.core import Path
+from scanspec.specs import Line, Spec
+
+# Here is the manual version of what we are trying to do
+
+# @strawberry.input
+# class LineInput(Line): ...
+
+
+# @strawberry.input
+# class ZipInput(Zip): ...
+
+
+# @strawberry.input(one_of=True)
+# class SpecInput:
+#     ...
+
+#     line: LineInput | None = strawberry.UNSET
+#     zip: ZipInput | None = strawberry.UNSET
+
+
+def generate_input_class() -> type[Any]:
+    # This will be our input class, we're going to fiddle with it
+    # throughout this function
+    class SpecInput: ...
+
+    # We want to go through all the possible scan specs, this isn't
+    # currently possible but can be implemented.
+    # Raise an issue for a helper function to get all possible scanspec
+    # types.
+    for spec_type in Spec.types:
+        # We make a strawberry input classs using the scanspec pydantic models
+        # This isn't possible because scanspec models are actually pydantic
+        # dataclasses. We should have a word with Tom about it and probably
+        # raise an issue on strawberry.
+        @strawberry.experimental.pydantic.input(all_fields=True, model=spec_type)
+        class InputClass: ...
+
+        # Renaming the class to LineInput, ZipInput etc. so the
+        # schema looks neater
+        InputClass.__name__ = spec_type.__name__ + "Input"
+
+        # Add a field to the class called line, zip etc. and make it
+        # strawberry.UNSET
+        setattr(SpecInput, spec_type.__name__, strawberry.UNSET)
+
+        # Set the type annotation to line | none, zip | none, etc.
+        # Strawberry will read this and graphqlify it.
+        SpecInput.__annotations__[spec_type.__name__] = InputClass | None
+
+    # This is just a programtic equivalent of
+    # @strawberry.input(one_of=True)
+    # class SpecInput:
+    #    ...
+    return strawberry.input(one_of=True)(SpecInput)
+
+
+SpecInput = generate_input_class()
+
+
+@strawberry.type
+class Query:
+    @strawberry.field
+    def validate(self, spec: SpecInput) -> str:
+        return validate_spec(spec)
+
+    @strawberry.field
+    def get_points(self, spec: Line, max_frames: int | None = 10000) -> PointsResponse:
+        """Calculate the frames present in the scan plus some metadata about the points.
+
+        Args:
+            spec: The specification of the scan
+            max_frames: The maximum number of frames the user wishes to receive
+        """
+
+        dims = spec.calculate()  # Grab dimensions from spec
+
+        path = Path(dims)  # Convert to a path
+
+        # TOTAL FRAMES
+        total_frames = len(path)  # Capture the total length of the path
+
+        # MAX FRAMES
+        # Limit the consumed data by the max_frames argument
+        if max_frames and (max_frames < len(path)):
+            # Cap the frames by the max limit
+            path = reduce_frames(dims, max_frames)
+        # WARNING: path object is consumed after this statement
+        chunk = path.consume(max_frames)
+
+        return PointsResponse(chunk, total_frames)
+
+
+schema = strawberry.Schema(Query)
+
+graphql_app = GraphQLRouter(schema, path="/", graphql_ide="apollo-sandbox")
+
+app = FastAPI()
+app.include_router(graphql_app, prefix="/graphql")
diff --git a/src/scanspec/schema/specs.py b/src/scanspec/schema/specs.py
new file mode 100644
index 00000000..1e629e06
--- /dev/null
+++ b/src/scanspec/schema/specs.py
@@ -0,0 +1,93 @@
+from __future__ import annotations
+
+from _collections_abc import Callable, Mapping
+from typing import Any
+
+import numpy as np
+import strawberry
+
+from scanspec.core import (
+    Axis,
+    Frames,
+    gap_between_frames,
+)
+
+
+@strawberry.type
+class PointsResponse:
+    """Information about the points provided by a spec."""
+
+    total_frames: int
+    returned_frames: int
+
+    def __init__(self, chunk: Frames[str], total_frames: int):
+        self.total_frames = total_frames
+        """The number of frames present across the entire spec"""
+        self.returned_frames = len(chunk)
+        """The number of frames returned by the getPoints query
+            (controlled by the max_points argument)"""
+        self._chunk = chunk
+
+
+@strawberry.interface
+class SpecInterface:
+    def serialize(self) -> Mapping[str, Any]:
+        """Serialize the spec to a dictionary."""
+        return "serialized"
+
+
+def _dimensions_from_indexes(
+    func: Callable[[np.ndarray], dict[Axis, np.ndarray]],
+    axes: list,
+    num: int,
+    bounds: bool,
+) -> list[Frames[Axis]]:
+    # Calc num midpoints (fences) from 0.5 .. num - 0.5
+    midpoints_calc = func(np.linspace(0.5, num - 0.5, num))
+    midpoints = {a: midpoints_calc[a] for a in axes}
+    if bounds:
+        # Calc num + 1 bounds (posts) from 0 .. num
+        bounds_calc = func(np.linspace(0, num, num + 1))
+        lower = {a: bounds_calc[a][:-1] for a in axes}
+        upper = {a: bounds_calc[a][1:] for a in axes}
+        # Points must have no gap as upper[a][i] == lower[a][i+1]
+        # because we initialized it to be that way
+        gap = np.zeros(num, dtype=np.bool_)
+        dimension = Frames(midpoints, lower, upper, gap)
+        # But calc the first point as difference between first
+        # and last
+        gap[0] = gap_between_frames(dimension, dimension)
+    else:
+        # Gap can be calculated in Dimension
+        dimension = Frames(midpoints)
+    return [dimension]
+
+
+@strawberry.input
+class Line(SpecInterface):
+    axis: str = strawberry.field(description="An identifier for what to move")
+    start: float = strawberry.field(
+        description="Midpoint of the first point of the line"
+    )
+    stop: float = strawberry.field(description="Midpoint of the last point of the line")
+    num: int = strawberry.field(description="Number of frames to produce")
+
+    def axes(self) -> list:
+        return [self.axis]
+
+    def _line_from_indexes(self, indexes: np.ndarray) -> dict[Axis, np.ndarray]:
+        if self.num == 1:
+            # Only one point, stop-start gives length of one point
+            step = self.stop - self.start
+        else:
+            # Multiple points, stop-start gives length of num-1 points
+            step = (self.stop - self.start) / (self.num - 1)
+        # self.start is the first centre point, but we need the lower bound
+        # of the first point as this is where the index array starts
+        first = self.start - step / 2
+        return {self.axis: indexes * step + first}
+
+    def calculate(self, bounds=True, nested=False) -> list[Frames[Axis]]:
+        return _dimensions_from_indexes(
+            self._line_from_indexes, self.axes(), self.num, bounds
+        )