How Do Solo Developers Run Multiple Agent-Driven Projects in Parallel?
For Solo developers and indie hackers building multiple projects · Based on Solmaz On-Demand Disposable Agent Orchestration Framework
// TL;DR
Solo developers and indie hackers can use the Solmaz framework to work on multiple projects simultaneously by running parallel channel workloads on Discord. Each Discord channel is bound to a separate agent harness session via ACPX, creating the equivalent of multiple full IDE sessions. This enables coding-on-the-go — dispatching tasks from your phone while commuting, monitoring progress across channels, and retrieving artefacts agents produce. The framework turns one developer into a small team by parallelising agent-driven work.
Why Can't Solo Developers Just Use One Agent at a Time?
Using one agent sequentially means waiting for each task to complete before starting the next. If you're building three side projects, a blog platform, an API wrapper, and a CLI tool, sequential execution wastes hours of idle time where you're waiting for the agent to finish before giving it the next task.
The Solmaz framework's parallel channel workloads pattern eliminates this. Run one agent per channel — Project-A, Project-B, Project-C — and all three progress simultaneously.
How Do You Set Up Parallel Channel Workloads on Discord?
1. Create dedicated channels: Name them by project or task — e.g. `blog-platform`, `api-wrapper`, `cli-tool`.
2. Bind each channel to a harness session: Use ACPX to connect each Discord channel to a separate agent harness (Codex, Claude Code, or OpenClaw). Each binding creates an on-demand disposable pod on your Kubernetes cluster.
3. Dispatch tasks: Send instructions to each channel. The agent in each channel works independently on its assigned project.
4. Monitor across channels: Check progress across 1-5 channels simultaneously. Each channel is effectively a full IDE session.
5. Retrieve artefacts: Use a dedicated channel to retrieve files the agent can't send directly through Discord — e.g. generated PDFs, compiled binaries, or images placed in `/tmp`.
This is Telegram Driven Development (TDD) — dispatching, monitoring, and iterating via messaging platform channels rather than traditional IDEs. It works from your phone, your tablet, or any device with Discord.
How Do You Manage State Across Multiple Projects?
Each project channel's agent pod has its own filesystem and GitHub access. Configure state synchronisation so each agent writes changes back to its project's repository. Since projects are independent, you don't need cross-project sync — just ensure each agent has read/write access to its own repo.
For projects that share code (e.g. a shared library used by multiple services), configure the rsync-style synchronisation layer across the relevant agent pods to prevent conflicting changes.
How Do You Handle the Cost of Running Multiple Kubernetes Pods?
For solo developers, cost matters. Strategies:
- Use spot instances: Configure your Kubernetes cluster to use cloud spot/preemptible instances for agent pods. Agent work is interruptible — if a pod dies, the Goal Operator restarts it.
- Set aggressive timeouts: Tear down idle pods after 10-15 minutes of inactivity.
- Right-size pods: Not every task needs a large pod. Use smaller resource requests for lightweight tasks.
- Use a managed Kubernetes service: Avoid the overhead of self-managed clusters. GKE Autopilot, EKS Fargate, or similar services handle node management for you.
The cost of running 3-5 agent pods for a few hours is typically less than $5 — far less than the value of parallelised development time.
Next step: Set up a Discord server with three project channels. Install ACPX and bind each channel to a Codex or Claude Code session. Dispatch your first parallel tasks and experience the compression of elapsed time firsthand.
// FREQUENTLY ASKED QUESTIONS
Do I need a full Kubernetes cluster to use this as a solo developer?
You need Kubernetes, but it doesn't have to be complex. Use a managed service like GKE Autopilot or EKS Fargate that handles node management. Alternatively, use a lightweight local cluster (k3s, minikube) for testing and a managed cloud cluster for production work. The Goal Operator and helm charts make deployment repeatable regardless of cluster size.
Can I use Slack or Telegram instead of Discord for parallel channel workloads?
Yes. ACPX binds to any supported communication platform through ACP adapters. Discord, Slack, Telegram, and Teams all work. Choose whichever platform you already use daily. The parallel channel pattern is identical — one channel per project, one agent per channel, all running concurrently. Discord is popular because it makes creating multiple channels effortless.
How do I handle tasks that need me to review output before the agent continues?
Build review checkpoints into your SOP workflow. The agent completes a step, emits structured JSON with its output, and pauses until you respond in the channel. This works naturally with the messaging platform model — you see the agent's output as a message, review it on your phone, and reply with approval or corrections. The agent resumes from the checkpoint with your feedback incorporated.