Skip to main content
Glama

save_draft

Save words the user wrote or pasted as a new draft, word for word, with no AI. If you are about to write the copy yourself, call draft_post instead: it writes from this person's real posts, voice and the client's brand, and nothing on this path does. If you still write it yourself, call get_brand_voice first and match it. When the saved copy doesn't sound like the author, the reply ends with a note saying what is off; offer revise_draft to rewrite it in their voice. IMPORTANT: This tool ONLY saves. Do NOT call queue_post or schedule_post unless explicitly asked. For YouTube, the post must have a video attached (via images with a video MIME type) before it can be scheduled or published. When saving a post that belongs to someone else on the workspace (for example moving a teammate's posts in), pass their name as author so it is filed under them. Pass tags to label the post the way it was labelled in the tool it came from. When the workspace has more than one account on a platform (two LinkedIn profiles, a profile and a company page), pass platform_account_ids so the draft publishes from the right one; it is kept on the draft and used when it is scheduled, queued or published. Pass first_comment to have a comment posted under the post once it is live (LinkedIn, Facebook, Instagram), or first_comments for several, each from its own profile. To give one channel different text (an X version of a LinkedIn post), pass channel_copies on the same post; don't save a second post. To tag a person or company page on LinkedIn ("tag Jane", "mention Acme"): call find_linkedin_mention first to get their exact name and urn, write "@" in the post text where the tag goes, and pass {name, urn} in linkedin_tags. A bare "@Name" in the text is never tagged by itself. Tags work in the LinkedIn post only, not in first comments. Pass penciled_at to pencil it in at a date and time. Drafts never post. A penciled time only holds the slot.

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": {
      "content": "<content>"
    },
    "name": "save_draft"
  }
}

Input Schema

NameRequiredDescriptionDefault
tagsNoTag names for the post, as they appear on the Content page (e.g. ["Launch", "Case study"]). Matched to the workspace's existing tags ignoring case; a name that doesn't exist yet is created if the user may manage tags, otherwise nothing is saved and the reply lists the tags that exist.
authorNoOptional: whose post this is. A teammate's name, or 'me'. Omit to save it as yours. Only people who can see everyone's posts may save as someone else.
imagesNoImages, a video, or one PDF to attach. For ANY image beyond a thumbnail, call upload_image FIRST to get a presigned S3 URL, upload via the returned curl command, then pass the s3_key here. Inline `data` is only for tiny images (≤50 KB); `file_path` is dev-only. A PDF posts as a LinkedIn document (a swipeable carousel): LinkedIn only, one per post, never alongside images or video.
contentYesThe post content to save
youtubeNoYouTube-specific options. Required platform inputs: a video must be attached via `images` (the upload tool accepts video MIME types). If `title` is omitted, it is derived from the first line of content (truncated to 100 chars).
platformsNoTarget platforms. Instagram needs an image or video on the post: upload one with `upload_image` (or make one with `generate_image`) and pass its s3_key in `images`.
workspaceNoOptional: the workspace (client) to answer from for this call only, by name (e.g. 'Acme') or numeric id, as returned by list_workspaces. Leave it out to use the current workspace. It does not change the current workspace; use select_workspace for that.
penciled_atNoOptional: pencil the draft in at a date and time, ISO 8601 (e.g. '2026-10-20T09:00'; no offset means the workspace's posting zone). Drafts never post. A penciled time only holds the slot. It stays a draft: nothing is scheduled or queued, and the queue skips that slot. A date with no time is refused. schedule_post with no `scheduled_at` books it at that time.
first_commentNoOptional: a comment DemandBird posts under the post, from the same profile, once the post is live (often a link or a call to action kept out of the post itself). LinkedIn, Facebook and Instagram only here. Stored verbatim. Pass {"text": ""} to remove it. Omit to leave it as it is.
linkedin_tagsNoOptional: real LinkedIn tags (people or company pages) in the LinkedIn post. Each is the {name, urn} `find_linkedin_mention` returned, or {url} with a LinkedIn profile or company page URL. Every tag's "@<name>" must be in the LinkedIn text, spelled exactly as returned. A tag that can't be resolved saves nothing. Tags already on the post stay while their "@Name" is still in the text. LinkedIn only; not in first comments.
channel_copiesNoOptional: a channel's own text, keyed by platform, e.g. {"twitter": "..."} for an X version of a LinkedIn post. That text publishes on that channel instead of the shared `content`; every other channel keeps what it had. Stored verbatim. A channel named here is added to the post's platforms. Pass "" for a channel to remove its copy (it then publishes the shared content). Use this, never a second post, to give one channel different text.
first_commentsNoOptional: up to 5 comments DemandBird posts under the post once it is live, in order, each on its own delay and, if `comment_as` is set, from its own profile or page (a teammate adding their thoughts). Replaces every first comment on the platforms it covers; [] removes them all. Stored verbatim. Use this or `first_comment`, not both. Omit to leave the comments as they are.
platform_account_idsNoOptional map of platform => connected_account_id (numeric id from `list_connected_accounts`): which account each platform publishes from. Example: {"linkedin": 42}. A platform named here is added to `platforms`. Each id must be a channel of this workspace and match the platform, or nothing is saved.

Schema Changelog

Changes observed during successful MCP inspections.

  1. Changed1 schema field changed
    • addedInput schema / properties / workspace
      Added value: +{
      +  "description": "Optional: the workspace (client) to answer from for this call only, by name (e.g. 'Acme') or numeric id, as returned by list_workspaces. Leave it out to use the current workspace. It does not change the current workspace; use select_workspace for that.",
      +  "type": "string"
      +}
  2. Changed4 schema fields changed
    • addedInput schema / properties / first_comment / properties / image
      Added value: +{
      +  "description": "Optional: one JPEG or PNG posted with the comment. LinkedIn and Facebook only (Instagram comments take text only, so the image is left off there). Upload it with `upload_image` first and pass its key as {\"s3_key\": \"...\"}. null removes the image. Omit to leave it as it is.",
      +  "properties": {
      +    "filename": {
      +      "description": "Optional: a name to show for it in DemandBird.",
      +      "type": "string"
      +    },
      +    "s3_key": {
      +      "description": "The s3_key `upload_image` returned.",
      +      "type": "string"
      +    }
      +  },
      +  "type": [
      +    "object",
      +    "null"
      +  ]
      +}
    • changedInput schema / properties / first_comment / properties / text / description
      Previous value: -"The comment, verbatim. LinkedIn allows 1250 characters, Facebook 2200, Instagram 2200. Empty removes the first comment."New value: +"The comment, verbatim. LinkedIn allows 1250 characters, Facebook 2200, Instagram 2200. Empty removes the first comment, unless `image` is set: then the comment is just the image."
    • addedInput schema / properties / first_comments / items / properties / image
      Added value: +{
      +  "description": "Optional: one JPEG or PNG posted with the comment. LinkedIn and Facebook only (Instagram comments take text only, so the image is left off there). Upload it with `upload_image` first and pass its key as {\"s3_key\": \"...\"}. null removes the image. Omit to leave it as it is.",
      +  "properties": {
      +    "filename": {
      +      "description": "Optional: a name to show for it in DemandBird.",
      +      "type": "string"
      +    },
      +    "s3_key": {
      +      "description": "The s3_key `upload_image` returned.",
      +      "type": "string"
      +    }
      +  },
      +  "type": [
      +    "object",
      +    "null"
      +  ]
      +}
    • changedInput schema / properties / first_comments / items / properties / text / description
      Previous value: -"The comment, verbatim."New value: +"The comment, verbatim. May be empty when `image` is set."
  3. Changed1 schema field changed
    • addedInput schema / properties / linkedin_tags
      Added value: +{
      +  "description": "Optional: real LinkedIn tags (people or company pages) in the LinkedIn post. Each is the {name, urn} `find_linkedin_mention` returned, or {url} with a LinkedIn profile or company page URL. Every tag's \"@<name>\" must be in the LinkedIn text, spelled exactly as returned. A tag that can't be resolved saves nothing. Tags already on the post stay while their \"@Name\" is still in the text. LinkedIn only; not in first comments.",
      +  "items": {
      +    "properties": {
      +      "name": {
      +        "description": "The name exactly as find_linkedin_mention returned it, e.g. \"Jane Doe\".",
      +        "type": "string"
      +      },
      +      "url": {
      +        "description": "Instead of name + urn: a LinkedIn profile (linkedin.com/in/...) or company page URL, looked up for you.",
      +        "type": "string"
      +      },
      +      "urn": {
      +        "description": "The exact urn find_linkedin_mention returned, e.g. \"urn:li:person:AbC123\" or \"urn:li:organization:1234\". Never make one up.",
      +        "type": "string"
      +      }
      +    },
      +    "type": "object"
      +  },
      +  "type": "array"
      +}
  4. First observed

TDQS

A4.6/5.0
Behavior5/5

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

Annotations only tell the agent this is a non-destructive, non-open-world write; the description adds the operational consequences: drafts never post, a penciled time only holds the slot, YouTube posts need a video attached via `images`, and a bare "@Name" is never tagged. It also flags failure modes such as an unresolvable tag saving nothing.

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

Conciseness3/5

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

The most decision-relevant content is front-loaded, but the body is a dense ~380-word run-on paragraph that mixes routing rules, parameter walkthroughs and platform caveats with no separation. Given 13 parameters some length is warranted, yet several clauses (first_comment vs first_comments, tag mechanics) largely restate the schema and could be tightened.

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 a 13-parameter tool with nested objects and no output schema, the description covers nearly every edge case an agent needs: platform requirements, tag resolution, account selection, and penciled-time semantics. The only gap is that it never says what a successful save returns (draft id or handle) beyond mentioning the optional voice-mismatch note.

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 coverage is 100%, so the baseline is 3; the description goes beyond it by explaining intent — `author` for filing a teammate's post under them, `platform_account_ids` for workspaces with multiple accounts on one platform, and `channel_copies` instead of saving a second post. The `linkedin_tags` workflow (call `find_linkedin_mention` first, then place "@<name>" in the text) is genuinely additive rather than a restatement.

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

Purpose5/5

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

Opens with a precise verb and resource: "Save words the user wrote or pasted as a new draft, word for word, with no AI." It immediately distinguishes itself from the closest sibling by naming `draft_post` and stating that nothing on this path does AI writing.

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

Usage Guidelines5/5

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

Explicit routing rules are given for each branch: use `draft_post` if you would write the copy yourself, call `get_brand_voice` if you still write it, offer `revise_draft` when the copy doesn't sound like the author. It also states a hard exclusion — do not call `queue_post` or `schedule_post` unless explicitly asked.

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

Try in Browser

Glama MCP Gateway

Add one secure layer between your agents and this server.

Resources