Current configured agent contract

Configured agent contract

A human creates the workspace key in Accounts. These 13 tools let the connected agent take setup from there.

Postdom shows the exact pinned MCP configuration once when the key is created. The public package is @postdom/mcp@0.3.0.

VERIFIED · 26 AUGUST 2026 · 13 CONFIGURED TOOLS

HUMAN OAUTHYour agent starts setup. You authorize the account.

Paste the configuration from Accounts. The agent checks workspace status, returns the social authorization link, discovers connected account IDs, and reads the Workspace Brief.

Callable surface

13 configured tools

get_workspace_statusStart here: read setup progress, the connection gate, accounts, policy, brief, and activity counts.

No inputs

list_accountsRead the workspace's connected destination IDs and current connection state.

No inputs

connect_accountOpen a platform authorization flow for a human to complete when workspace connections are unlocked.

platform: instagram | tiktok | youtube | linkedin | facebook | snapchat | twitter | threads

get_briefRead the current workspace-owned brand guidance before planning or writing content.

No inputs

get_digestRead the latest completed workspace week with populations, cadence endings, and coverage gaps.

No inputs

upload_mediaCreate a short-lived direct PUT contract for finished video bytes.

content_type · size_bytes · platforms

get_mediaVerify an uploaded media handle before publishing it.

media_handle

submit_planPropose a bounded publication plan for human approval.

account_ids · title · objective · starts_at · ends_at · max_posts · intent · brief_version? · agent_identity? · idempotency_key?

get_planRead approval status and structured human feedback.

plan_id

publish_videoSubmit finished media now or at an exact UTC time; approval depends on trust and plan authorization.

account_ids · exactly one of video_url or media_handle · caption · intent · agent_identity? · publish_at? · plan_id? · idempotency_key?

get_publishRead publication and destination state.

post_id

get_performanceRead normalized post or account measurements.

Exactly one of post_id or account_id · account window: 7d | 30d

get_best_postsRank account posts by one evidence-backed metric; nulls remain excluded with reasons.

account_id · metric · window: 7d | 30d

First hour

Status before action

Call get_workspace_status first. If connections are locked, return the exact gate detail. If unlocked, use connect_account only to create a human authorization handoff, then call list_accounts until the connected destination appears.

Never guess an account ID or enter a human's social credentials.

Workspace guidance

Read before writing

Call get_brief before planning or writing content. Before submit_plan, read get_digest and the prior plan outcome when one exists. These describe what happened; they do not recommend what to do.

Follow the brief and attach its version as brief_version. Hard schedules, caps, and publishing windows remain in Policy.

Scheduling boundary

Scheduling capability

upload_media returns a short-lived direct PUT contract; get_media verifies the handle before publish_video schedules it inside account policy or an approved plan.

Agents provide exactly one video_url or stored media_handle, attach the approved plan’s id when required, and never guess a timezone or unsupported platform setting.

Authority boundary

Human authority

The agent cannot promote itself.

Human: create workspace · create or revoke agent keys · complete social authorization · approve plans · set trust · pause · handle exceptions

Agent: read setup status · initiate authorization handoff · discover destination IDs · read workspace guidance · upload finished media · propose plans · submit supported work · read state · read measurements

Never: expose credentials or signed upload URLs · enter social credentials · approve itself · expand trust · invent schema fields