Skip to main content
Glama
okenjioxx

Roblox Executor MCP Server

by okenjioxx

Stream bounded event-driven world changes

world-delta
Destructive

Capture live Roblox client world changes via RBXScriptConnections, then poll cursor-ordered deltas for additions, removals, camera, character, and tool events without modifying gameplay Instances.

Instructions

WRITES LIVE CLIENT OBSERVER STATE — installs RBXScriptConnections and a bounded getgenv registry, but never modifies gameplay Instances. action='start' subscribes to Workspace and PlayerGui descendant additions/removals, character spawn/removal and equipped Tool changes, CurrentCamera replacement and selected camera properties, Backpack Tool changes, plus explicitly requested Instance properties. action='poll' returns cursor-ordered deltas; action='status' reports observer health; action='stop' disconnects every connection. Events are filtered before storage, repeated noise is coalesced/throttled, the ring buffer reports every capacity eviction, and a fixed TTL runs the same cleanup path automatically. It uses Roblox signals only: no RenderStepped, per-frame polling, or world scans. Signature: { action: "start" | "poll" | "status" | "stop", observerId: string?, cursor: any?, limit: any?, maxEvents: any?, ttlSeconds: any?, throttleMs: any?, coalesceWindowMs: any?, filters: any?, cameraProperties: any?, watchedProperties: any?, threadContext: number? }. Phase: observe; cost=medium; idempotency=contextual-write. Requires: active-client, explicit-mutation-approval, getgenv capability, a bounded observation goal and relevant source filters. Capabilities: getgenv. Produces: grounded-evidence, event-driven-world-deltas, monotonic-cursor, buffer-gap-and-drop-accounting, observer-health-and-expiry, coalescing-and-throttle-statistics. Verify with: assert-state, observe-world. Safety: MUTATING; writes live game/client state, installs bounded RBXScriptConnections in the active client, stores bounded observer metadata and events in getgenv.__mcp_world_delta, schedules one TTL cleanup callback; no gameplay Instance is written. On failure: inspect tool-schema for exact fields, defaults, constraints, and an invocation example.

Input Schema

TableJSON Schema
NameRequiredDescriptionDefault
limitNoMaximum events returned by one poll; hasMore indicates another page.
actionYesstart installs a bounded observer, poll reads events after cursor, status reports health without consuming events, and stop disconnects every observer connection.
cursorNoPoll returns events whose monotonically increasing cursor is greater than this value. Always persist the returned nextCursor for the next poll.
filtersNoStart-only server-side filters, applied before an event consumes buffer capacity.
maxEventsNoStart-only ring-buffer capacity. Older events are evicted at the cap and reported through droppedEvents, droppedThrough, and cursorGap.
observerIdNoObserver id returned by start. For poll/status/stop, omission selects the newest active observer owned by this MCP session.
throttleMsNoMinimum interval for storing repeated events with the same semantic key after their coalescing window.
ttlSecondsNoStart-only fixed lifetime. A delayed cleanup retires the observer and disconnects every connection even if the AI never calls stop.
threadContextNoOptional Roblox thread identity for this call; omit it to use the server default.
cameraPropertiesNoCamera properties watched with GetPropertyChangedSignal. Add CFrame or Focus only when needed; throttling protects high-frequency properties.
coalesceWindowMsNoWindow in which identical source/kind/path/property events update one record and increment repeatCount.
watchedPropertiesNoAdditional explicit Instance/property subscriptions. At most 16 targets and 64 aggregate property connections are allowed.

Schema Changelog

Changes observed during successful MCP inspections.

  1. First observedv2.0.0-spies.2

TDQS

A4.3/5.0
Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Annotations already declare destructiveHint=true and readOnlyHint=false, but the description adds substantial context beyond them: it clarifies exactly what is written (client observer state, getgenv.__mcp_world_delta) and what is NOT ('no gameplay Instance is written'), plus TTL cleanup, ring-buffer eviction accounting, and coalescing/throttling behavior. This is a rich, non-contradictory disclosure.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The core behavior is front-loaded in the first sentence before the action enumeration, and the structured metadata (Phase/cost/idempotency/Requires/Capabilities/Safety) is scannable. It is dense and slightly redundant with the inline signature block, but every section carries usable information.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

For a 12-parameter, nested-schema tool with no output schema, the description covers what it produces (grounded-evidence, deltas, cursor, health), verification tools (assert-state, observe-world), safety, and failure guidance ('inspect tool-schema for exact fields...'). Nothing an agent needs to call it correctly appears missing.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema description coverage is 100%, so the baseline is 3. The description adds meaning beyond the schema by spelling out what action='start' subscribes to, how poll returns 'cursor-ordered deltas', and how status/stop behave, giving the agent inline semantic context for the key action parameter.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose4/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description states a specific verb+resource ('installs RBXScriptConnections and a bounded getgenv registry' that 'streams event-driven world changes') and enumerates the four actions. It hints at differentiation from scanning siblings ('no RenderStepped, per-frame polling, or world scans') but never names an alternative tool explicitly, so it falls just short of a 5.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines4/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

It gives clear context for invocation ('Requires: active-client, explicit-mutation-approval, getgenv capability, a bounded observation goal and relevant source filters') and explains each action's purpose. It does not, however, tell the agent when to choose this over observe-world, watch-value, or watch-property-changes, so exclusions are absent.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Deploy Server

Other Tools