Autonomous creator workflow runtime

BeaconCLIrunscreatoragents.

A local TypeScript CLI/runtime for coordinating autonomous social media creator systems: trend discovery, strategy, scripts, media, editing, rendering, publishing, community replies, analytics, and optimization feedback.

10

pipeline stages

9

agent roles

local

JSON memory

BeaconCLI pixel antenna logo on a light purple background

npm run waypoint -- run gaming

local run - deterministic adapters - inspectable output

Creator-agent run

One objective, many coordinated agents.

local SDK

Trend

detect gaming clip velocity

complete

Script

write hook, beats, captions

complete

Render

compose Remotion-style output

running

Publish

prepare TikTok + YouTube packet

queued

Analytics

score retention and comments

waiting

Runtime model

Autonomoussocialworkflowsneedcoordination,notanotherpromptbox.

BeaconCLI keeps each stage explicit: who owned it, what typed input crossed the boundary, which artifact was produced, which provider action was attempted, what memory changed, and how analytics influence the next run.

Social providers

Adapter boundaries for Instagram, TikTok, YouTube, X, Telegram, Meta Graph, publish packets, scheduling, comments, replies, and analytics.

Media providers

Structured composition, image, voice, render, and Remotion-style outputs that can be swapped for real integrations later.

Local persistence

Pipeline runs, artifacts, memory commits, and provider actions are persisted as inspectable JSON under local data directories.

Gitlawb memory

Versioned memory records with DID-style agent IDs, parent commits, changed files, timestamps, and run metadata.

Thesamplepipelinecoversthefullcreatorloop.

A rough creator objective becomes a repeatable staged run with typed state, generated artifacts, provider boundaries, community feedback, analytics review, and optimization recommendations for the next cycle.

01trend_discovery
02strategy_planning
03scriptwriting
04media_generation
05video_editing
06remotion_render
07platform_publish
08community_engagement
09analytics_review
10optimization_loop

Agent contract

Specialistcreatoragentswithstructuredoutputs.

TrendDiscoveryAgent

finds platform velocity and creator opportunities

StrategyAgent

turns trends into audience, hook, and channel plans

ScriptwritingAgent

writes structured scripts, beats, captions, and CTAs

VideoGenerationAgent

creates media artifacts for the selected concept

EditingAgent

assembles cuts, timing, captions, and variant metadata

PublishingAgent

prepares platform actions for social distribution

CommunityAgent

reads comments and drafts creator-native replies

AnalyticsAgent

scores retention, saves, comments, and performance

OptimizationAgent

feeds recommendations into the next run

Run, inspect, and clean local creator workflows.

BeaconCLI follows the Waypoint SDK local-first model: no API keys required, deterministic provider mocks, reproducible outputs, and every run inspectable from local JSON.

Autonomous creator accounts

AI influencer networks

Agency-run social agents

Gaming short-form content systems

Lifestyle creator workflow automation

Cross-platform publishing experiments

Analytics-driven content iteration

Local-first runtime prototyping

~/beacon

$npm install
$npm run waypoint -- run gaming
$npm run waypoint -- inspect <runId>
$npm run waypoint -- memory
run complete - 10 stages - 14 artifacts - 3 provider actions - 5 memory commits

Local data layout

persisted creator workflow state

Run state

stage status, assigned agent, schema names, timestamps

Artifacts

scripts, media records, captions, render metadata, provider outputs

Provider actions

publish attempts, scheduled posts, analytics fetches, comment replies

Memory commits

strategy changes, run summaries, optimization notes, parent links

Analytics

scores, recommendations, retention signals, platform feedback

Inspection

CLI commands for run detail, memory history, and cleanup

Gitlawb-style memory gives every agent run history.

Memory commits record the author agent, message, timestamp, parent commit, files changed, and run metadata. That makes strategy updates, optimization notes, and execution history inspectable after the content is produced.

did:gitlawb:agent:beacon:trend-discovery
parentCommitId: 91e7bb
filesChanged: strategy.json, analytics.json

Analytics feed the next creative decision.

The runtime does not stop at publishing. It fetches analytics-shaped data, reads comments, scores performance, and passes recommendations into the optimization loop so the next content run can adapt.

Trend discovery

Track what the next content run should react to.

Community replies

Draft replies and extract audience signals.

Render outputs

Keep final media tied to the run that produced it.

Optimization

Convert analytics into memory and next-run strategy.

Local by design while contracts evolve.

The current runtime uses deterministic local adapters rather than real model calls, social APIs, renderers, or Gitlawb infrastructure. That keeps provider boundaries explicit, output reproducible, and the orchestration contract testable.

Roadmap

Stronger zod schemas for every stage boundary

Retry policies and resumable pipeline runs

Pluggable agent registry

Real provider adapters behind the existing contracts

Queue-backed execution

First-class artifact store

Human review gates

Policy checks for platform safety and brand constraints

Memory backend integration behind the current contract

Render job abstraction for Remotion and alternate renderers

Private whitelist

Getearlyaccesstothecreator-agentruntime.

Join the BeaconCLI whitelist for teams experimenting with autonomous creator accounts, social-agent workflows, platform publishing adapters, and analytics-driven content loops.

Creator agents

Local JSON runs

Provider adapters

Gitlawb memory

This preview form only changes state locally and does not send data anywhere yet.

Open source

Build the creator-agent runtime before real provider lock-in.

BeaconCLI keeps the Waypoint-style creator workflow contract visible and testable first, then leaves room for real social, media, memory, and rendering providers behind stable adapters.