premiere-pro-mcp

Overview Schema Source Related Servers Discussions Admin

Manage Project Context

manage_project_context

Destructive

Capture, enrich, import, inspect, or clear a durable local Premiere project-context index for revision-bound sequence metadata and caller-approved editorial evidence without re-analyzing media.

Instructions

Capture, enrich, import revision-bound editorial evidence, inspect, or clear a durable local Premiere project-context index. Capture stores bounded active-sequence/source metadata; local enrichment and evidence import add caller-approved transcripts, speaker labels, shots, audio observations, notes, or opaque frame references without re-analyzing media. Never include secrets or unrelated customer data.

Tool call template

Replace each placeholder with a value matching the input schema. Only required parameters are included; add optional parameters as needed.

{
  "id": 1,
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "arguments": {
      "action": "<action>"
    },
    "name": "manage_project_context"
  }
}

Input Schema

Name	Required	Description
`action`	Yes	Capture active context, enrich it, import revision-bound editorial evidence, inspect status, or clear one local project index.
`records`	No	For enrich, up to 512 transcript, shot, audio, or note records
`replace`	No	For enrich, replace prior enrichments while retaining captured core records
`evidence`	No	For import_evidence, up to 512 caller-approved records with strict source/timeline revision guards. The server does not read referenced frame files, invoke analysis, or contact a provider.
`project_id`	No	Project context ID returned by capture; optional for status to list projects
`idempotency_key`	No	Optional client-chosen key (1-128 chars of A-Z a-z 0-9 . _ : -) for this intended edit. Pass a fresh key per edit and reuse it only when retrying that same edit, for example after a client-side timeout. A repeat with the same tool, key and arguments is not run again: while the first call is running it fails with IDEMPOTENCY_IN_PROGRESS (poll get_last_operations); after it finished it returns the original outcome marked replayed. A definite failure with nothing applied (receipt outcome not_applied) may be retried with the same key. Reusing a key with different arguments fails with IDEMPOTENCY_KEY_REUSED. Scoped to this MCP session (the stdio server process, or the HTTP server process when Streamable HTTP is stateless per request) and held in memory only (200 keys, 1 hour); it does not survive a restart or reconnect. Read-only invocations are never deduplicated.
`expected_cep_target`	No	Optional reviewed CEP sequence/clip identity and exact tick ranges. Checked before every generated command; UXP refuses this CEP-only guard.
`expected_project_path`	No	Optional absolute saved-project path. Every generated CEP command refuses before its body when the active project differs or is unsaved. UXP commands refuse this CEP-only guard; local-only tools do not inspect Premiere. Omit to keep existing behavior.

Output Schema

Name	Required	Description
`ok`	Yes	Whether the tool completed successfully.
`data`	No	Tool-specific result data when ok is true; on failure, diagnostic detail when the tool provides it.
`tool`	Yes	The registered MCP tool name.
`error`	No	Failure detail when ok is false.
`receipt`	No	Versioned common mutation receipt; success never implies rendered verification.
`idempotency`	No	Present when idempotency_key deduplicated this call: replayed is true when the original outcome was returned instead of running again.

Schema Changelog

Changes observed during successful MCP inspections.

First observedv1.30.0Nov 14, 2026

TDQS

A3.8/5.0

Behavior3/5

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

Annotations already declare destructiveHint=true, openWorldHint=false, and readOnlyHint=false, so the safety profile is supplied. The description does add genuinely useful behavior: the index is local and durable, it stores only bounded metadata, and enrichment never triggers media re-analysis or provider calls. However, for a destructiveHint=true tool it never explains the blast radius of 'clear' beyond 'one local project index', nor mentions revision-guard rejection behavior that dominates the schema.

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

Conciseness5/5

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

Three sentences, front-loaded with the operations, then storage contents, then the privacy guardrail. Every clause carries information an agent needs (local, durable, bounded, caller-approved, no media re-analysis, no secrets), and nothing is repeated from the schema.

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

Completeness4/5

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

For an eight-parameter tool with nested objects, enum actions, and an output schema, the description covers purpose, per-action intent, locality, and a data-hygiene constraint. Because the output schema exists, the absence of return-value detail is acceptable, and the schema itself carries the tick/revision guard semantics. It loses a point only for not stating the capture-then-enrich ordering prerequisite that an agent must satisfy.

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

Parameters3/5

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

Schema description coverage is 100%, so the schema already carries the full parameter burden, including the revision-guard requirements and idempotency_key semantics. The description reinforces the action vocabulary and lists the record kinds (transcripts, speaker labels, shots, audio observations, notes, frame references) that map to the evidence enum. With a schema this rich, baseline 3 is appropriate since little is added beyond restatement.

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 names a specific resource (a durable local Premiere project-context index) and enumerates the five operations performed against it: capture, enrich, import evidence, inspect, clear. That is far more precise than the generic name suggests. It stops short of 5 because it never contrasts with the obvious read counterpart search_project_context, leaving the agent to infer which tool answers a query versus which one mutates the index.

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?

Each action is given a use condition: capture for active-sequence/source metadata, enrich and import_evidence for caller-supplied editorial data, and the phrase 'without re-analyzing media' implicitly steers the agent away from analysis tools toward this one. There is no explicit when-not guidance or named alternative, and no statement that project_id must already exist from capture before enrich/import work, so it falls short of a full 5.

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

Deploy Server