MoltbookPilot 👽🟢

Neon-themed ASP.NET Core Razor Pages “ops console” for running a local LM Studio agent that can register/verify on Moltbook, read the feed, draft posts, and publish — while persisting agent state in SQL Server.

Local-first: This app talks to your locally-running LM Studio server (OpenAI-compatible /v1 endpoints) and keeps secrets out of source control using User Secrets.

---

What it does

LM Studio tools

• Connectivity checks
  • Ping GET /v1/models
  • Send prompts via POST /v1/chat/completions

Moltbook onboarding + state

• Join / claim flow
  • “Join Moltbook” registers an agent and returns a claim link
  • Saves agent state (handle, claim URL, API key, heartbeat timestamps) in SQL Server
• State dashboard
  • View stored claim URL / masked API key
  • Refresh state from the backend

Heartbeats

• Manual heartbeat endpoint
  • POST /api/moltbook/heartbeat runs the heartbeat runner once
• Background hosted service
  • Periodic timer calls heartbeat runner (runner enforces the 4+ hour cadence)

Compose from Feed (new)

• Read N recent posts from a submolt (e.g. m/general)
• Mix them with your User Context and generate a draft
• Edit the draft in the UI
• Publish the draft to Moltbook

---

Architecture (high level)

• Razor Pages UI: Pages/Index.cshtml (Neon Alien Ops Console)
• LM Studio client: LmStudioClient (OpenAI-compatible Chat Completions)
• Tool runtime: AgentTools (safe HTTP GET/POST to moltbook.com only)
• Persistence: MoltbookAgentState + MoltbookDbContext + MoltbookStateStore
• Services
  • MoltbookJoinService (onboarding)
  • MoltbookHeartbeatRunner + MoltbookHeartbeatHostedService (heartbeat)
  • MoltbookComposeService (feed → draft → publish)
• Endpoints (typical)
  • GET /api/health
  • POST /api/agent/think
  • POST /api/moltbook/join
  • GET /api/moltbook/state
  • POST /api/moltbook/heartbeat
  • POST /api/moltbook/compose/preview (read posts + draft)
  • POST /api/moltbook/compose/publish (publish draft)

---

Prerequisites

• .NET SDK matching the project’s TargetFramework (e.g., net9.0 / net10.0)
• LM Studio installed and running with:
  • A model downloaded and loaded
  • Local Server enabled (commonly http://localhost:1234)
• SQL Server (any of):
  • SQL Server Express (.\SQLEXPRESS)
  • LocalDB ((localdb)\MSSQLLocalDB)
  • Full SQL Server instance
• (Optional) SSMS for inspecting the DB

---

Setup

1) Clone and restore

git clone <YOUR_REPO_URL>
cd MoltbookPilot
dotnet restore

2) Configure User Secrets (required)

This project uses User Secrets for local development so secrets do not land in appsettings.json or Git.

From the project directory (the folder containing the .csproj):

dotnet user-secrets init

Set your SQL connection string:

SQL Express example

dotnet user-secrets set "ConnectionStrings:MoltbookPilotDb" "Server=.\SQLEXPRESS;Database=MoltbookDb;Trusted_Connection=True;MultipleActiveResultSets=true;TrustServerCertificate=True"

LocalDB example

dotnet user-secrets set "ConnectionStrings:MoltbookPilotDb" "Server=(localdb)\MSSQLLocalDB;Database=MoltbookDb;Trusted_Connection=True;MultipleActiveResultSets=true;TrustServerCertificate=True"

3) Optional config (recommended)

dotnet user-secrets set "Agent:Model" "qwen/qwen3-coder-30b"
dotnet user-secrets set "Moltbook:BaseUrl" "https://www.moltbook.com"

Optional API path overrides (usually you do NOT need these unless Moltbook changes):

dotnet user-secrets set "Moltbook:FeedPath" "/api/v1/feed?limit={limit}"
dotnet user-secrets set "Moltbook:SubmoltFeedPath" "/api/v1/posts?submolt={submolt}&limit={limit}"
dotnet user-secrets set "Moltbook:CreatePostPath" "/api/v1/posts"

Tip: Prefer https://www.moltbook.com to avoid redirects that can strip Authorization headers.

---

Database setup

Option A (recommended): EF Core migrations

dotnet ef migrations add InitMoltbookState -c MoltbookDbContext
dotnet ef database update -c MoltbookDbContext

Option B: Create the table manually (SSMS)

CREATE TABLE dbo.MoltbookAgentState (
    Id               INT IDENTITY(1,1) PRIMARY KEY,
    AgentHandle      NVARCHAR(100) NULL,
    ClaimUrl         NVARCHAR(400) NULL,
    AgentApiKey      NVARCHAR(200) NULL,
    LastHeartbeatUtc DATETIME2(0) NULL,
    CreatedUtc       DATETIME2(0) NOT NULL DEFAULT SYSUTCDATETIME(),
    UpdatedUtc       DATETIME2(0) NOT NULL DEFAULT SYSUTCDATETIME()
);

---

Run the app

dotnet run

Open the URL printed in the console (usually https://localhost:xxxx).

---

Using the console

LM Studio

• Ping LM Studio → checks /v1/models
• Test a Prompt → sends prompt to /v1/chat/completions

Moltbook onboarding

1. Click Join Moltbook
2. The app should return:
   • a claim URL
   • an API key
3. Open the claim URL and complete verification
4. Return to the console and confirm your Moltbook state shows the key (masked)

Heartbeats

• Click Run Heartbeat
• The runner will fetch Moltbook heartbeat instructions and execute them (subject to cadence enforcement)

Compose from Feed (new)

1. Enter a submolt like m/general (or just general)
2. Choose how many posts to read (e.g. 15)
3. Add User prompt / context (tone, facts, what to focus on)
4. Click Read + Draft
5. Edit the draft (title on first line, blank line, then content)
6. Click Post Draft

---

Notes / gotchas

• Submolt formats
  • UI can accept general, m/general, or /m/general
  • API calls should use the slug (general) when filtering/creating posts
• Rate limits
  • Posting may return 429 Too Many Requests if you post too frequently. Handle gracefully and retry later.
• Prompt format
  • Drafting expects:
    • First line = title only (no "TITLE:" prefix, no markdown)
    • Blank line
    • Body/content

---

Security notes

• Never commit
  • Moltbook API keys
  • Connection strings with credentials
• Use User Secrets locally; use environment variables/secret store in production.

---

License

Licensed under the Apache License, Version 2.0. See the LICENSE file for details.

MoltbookPilot
Copyright (c) 2026 Gabriel Vogt