Distribution infrastructure · API and MCP coming in V1B

Distribution, as a capability your stack can call.

Brands, the network, campaign runs, evidence and results live on one system. Today your team works it with us directly. In V1B your agents will work it through typed, policy-bound tools, and you can request early access now. Agents will never get unrestricted access to your money, and autonomous execution will be opt-in, one action class at a time.

Request API access

The mascot works at a small console whose threads run out to the network; one node glows yellow, waiting for approval

In one paragraph

distribution.wtf is making distribution a capability software can call. In V1B a company's agent will be able to create an intent, resolve the audience, request a plan and inventory, create a campaign, route it for approval, execute approved actions, follow status, and retrieve evidence and results. Spend, new terms and publishing will stop for human approval unless policy allows them. Early access requests are open now.

Per our V1 product spec, the agent surface is 11 typed tools, from create_intent to stop, with approval on spend, terms and publishing.

The system

One object model, for people and for agents.

A brand's intent becomes a plan, the plan becomes a campaign run, the run produces evidence, and the evidence rolls up into results. The network supplies and fulfils. Your team works on these objects today; in V1B your agents will work on the same ones.

Agent layer · REST · MCP · webhookscalls every object below, inside your policy

01
Brand
company_id
product, ICP, policies
02
Intent
intent_id
audience, objective, budget
03
Plan
plan_id
nodes, allocation, deliverables
04
Campaign run
run_id
state, approvals, actions
05
Evidence
evidence_id
artifact, timestamp, verified
06
Results
outcome_id
delivery + outcome signals

Distribution networksupplies the plan and fulfils the run

Creators
Podcasts
Newsletters
Events
Communities
Media
Partners

A run, from the agent's side

What your agent actually sees.

mcp · distribution.*Sample run

distribution.create_intent({ product: "Series A", audience: "US AI engineers", budget: 50000 })→ intent_id: int_7q2k
distribution.resolve_audience({ intent_id })→ 5 distribution categories
distribution.plan({ intent_id, max_spend: 50000 })→ plan_9c1f · ranked plan + allocation
distribution.inventory({ plan_id, window: Oct 10..24 })→ available nodes + terms
distribution.create_campaign({ plan_id, assets: approved })→ cmp_2hx8 · state: approval
distribution.approve({ campaign_id, token })→ waiting on human: jane@company
distribution.execute({ campaign_id })→ run_5ka1 · executing
distribution.status({ run_id })→ per-placement state + blockers
distribution.evidence({ campaign_id })→ proof bundle · one per placement
distribution.results({ campaign_id })→ delivery + outcome signals
distribution.stop({ campaign_id, reason })→ available at any state

Policy

spend limit: $50,000 per campaign
channels: creator, podcast, newsletter, event
geography: US only
needs approval: spend, new terms, publishing
auto-allowed: planning, inventory, status

Campaign state

draft
planned
approval
approved
executing
live
complete

Every run keeps its state. It can pause for approval, resume, block on a person, or stop, and every action lands in an audit log.

Tool surface

Eleven tools. Narrow, typed, policy-bound.

Spend, non-standard terms, external publishing and live changes stop for approval unless your policy pre-authorises them.

Tool	Input	Output
distribution.create_intent	product, audience, objective, budget, dates, constraints	intent_id
distribution.resolve_audience	intent_id or audience profile	audience profile + distribution categories
distribution.plan	intent_id + constraints	ranked plan + budget allocation
distribution.inventory	audience, filter, date, budget	available nodes + terms
distribution.create_campaign	approved plan + assets	campaign_id + approval state
distribution.approve	campaign_id + approval token	execution state
distribution.execute	campaign_id	run_id + actions
distribution.status	campaign_id or run_id	current state + blockers
distribution.evidence	campaign_id	proof bundle
distribution.results	campaign_id	delivery + outcome metrics
distribution.stop	campaign_id	stopped state + reason

The agent layer

Built for machines, controlled by people.

The same campaign, states and evidence your team sees in the workspace, exposed to your software.

REST API
in build
Stable machine interface for any internal system or agent
MCP server
in build
Distribution capabilities as tools for compatible agents
Webhooks
in build
Status and evidence pushed back to your environment
Policy engine
in build
Spend, channel, geography and approval constraints
Run state
in build
Pause and resume multi-step jobs without losing context
Agent identity
in build
Know which client system started a run
Audit log
in build
Immutable action and decision history

Example run

A brand-side agent ships a launch.

“We are launching Product X. Target US consumers aged 25 to 40 interested in personal productivity. Budget $50K. Launch window October 10 to 24. Use approved creative only. Favour podcast, YouTube, newsletter and creator distribution. Do not spend outside the approved budget.”

01The agent calls create_intent.
02distribution.wtf resolves the audience and returns the relevant categories.
03The planner assembles inventory and a $50K draft allocation.
04Your policy engine checks budget, channels and approved assets.
05Only the approvals your policy requires go to a person.
06Execution coordinates bookings, assets and publishing.
07Every placement records evidence.
08Webhooks notify the agent on each state change.
09At close, the agent retrieves results and the evidence bundle.

Early access

Connect an agent.

API and MCP access opens to early design partners first. Tell us what your agent does today.