Analysis updated 2026-06-24
Fork the template and run a free Solana watcher and writer swarm on GitHub Actions
Add a custom skill by dropping a SKILL.md file under skills and wiring it to an agent
Chain skills into a morning pipeline that briefs, scans arxiv, and posts a digest
Plug Guardian into a Telegram channel as a kill switch over an autonomous trading bot
| stoaaadev/stoa | moonpiesheldon1337/mobsf-fail-app | clipboardhealth/groundcrew | |
|---|---|---|---|
| Stars | 30 | 30 | 31 |
| Language | TypeScript | TypeScript | TypeScript |
| Setup difficulty | hard | moderate | moderate |
| Complexity | 4/5 | 3/5 | 4/5 |
| Audience | developer | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
Executor can sign real Solana transactions, so wallet keys and API secrets must be configured carefully before enabling cron.
stoa is a TypeScript framework for running a coordinated swarm of seven AI agents that work together on autonomous tasks, mainly focused on Solana token trading, research, and content generation. The whole system is designed to run for free on GitHub Actions, with state stored as git commits and individual skills written as markdown prompt files rather than code. The seven agents each have a clear role. Scout watches Solana for token movements, volume spikes, and whale transactions. Analyst scores incoming signals against a confidence threshold. Executor places trades through Jupiter and supports DCA and stop-loss. Guardian is the risk manager and has veto power: a single halt message from it freezes the entire swarm. Researcher reads arxiv papers, protocol docs, and exploit post-mortems. Writer produces newsletters, digests, and social posts. Ops handles CI, dependency audits, PR review, and security scans on the repo itself. A skill is a markdown file under skills/your-skill/SKILL.md, with a YAML frontmatter that names the skill, picks the owning agent, optionally schedules it via cron, and describes its objective, steps, output, and exit codes. To add a new capability you drop a markdown file in the right place and reference it from the agent's config. Skills can be chained into pipelines with explicit dependencies, for example a morning pipeline that runs the morning brief, an arxiv scan, an analysis step, and then a daily digest. Getting started: fork the template repo, set ANTHROPIC_API_KEY and optionally Solana keys as repository secrets, enable the tick, agent, and messages workflows, and the swarm starts at the next cron tick. There is also a Next.js dashboard for monitoring the mesh and a CLI with commands like stoa dispatch, stoa health, stoa cost, and stoa gateway. Telegram, Discord, and Slack messaging is built in, and the LLM layer can fail over between Claude, OpenAI, and Gemini. The project is MIT-licensed and ships with pipeline tests and Solana devnet integration tests. Because the executor agent can sign on-chain Solana transactions when given a wallet key, anyone deploying this should treat it as an autonomous trading bot.
TypeScript framework that runs a seven-agent swarm on GitHub Actions for Solana trading, research, and content, with skills written as markdown files and state stored in git.
Mainly TypeScript. The stack also includes TypeScript, Next.js, GitHubActions.
MIT license: use freely, including commercially, as long as you keep the copyright notice.
Setup difficulty is rated hard, with roughly 1day+ to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Verify against the repo before relying on details.