alexcinovoj.devby TechTide AI
Contact
Back to resources
Open-source project

Swarm 357

357 Claude agent roles across six business layers, with portable memory, enforced cost caps, a bash policy gate, and durable checkpoints.

Project details

Classification
Open-source project
Repository
TechTideOhio/swarm-357
Language
Python
License
Apache-2.0
Status
Active development
Tests
4 test notes
Last verified
2026-08-02

What it is

I built Swarm 357 around a library of agent role definitions. The runtime checks spending limits and shell permissions, and stores checkpoints for recovery after interruption. Read the role files, evaluation directory, and status notes to inspect the implementation and unfinished work.

Live product

Architecture

  • Role library of 357 agent definitions grouped into six business layers.
  • Portable Memvid memory so context survives process boundaries.
  • Enforced per-run and per-agent cost caps.
  • Bash policy gate mediates shell access rather than granting it wholesale.
  • Durable checkpoints in the database directory allow resume after interruption.
  • FastAPI service surface, Docker image, Railway deployment target.

Tests and gates

  • Four GitHub Actions workflows: ci, codeql, publish, and publish-image.
  • An evals directory is checked in alongside the runtime.
  • gitleaks configuration guards against credential leakage in commits.
  • STATUS.md and ROADMAP.md record what is complete and what is not.

Limits

  • Role count is a capacity figure, not a performance claim. Concurrency in a given run depends on budget and host.
  • Cost caps bound spend; they do not bound wall-clock time.
  • Memory portability is scoped to the Memvid format used by the runtime.

Source files and examples

Related methods

Acceptance tests for agent work

Executable checks test the task and detect regressions.

Evaluation tests for releases

A fixed evaluation suite checks release requirements.

MCP tool permissions

The server checks which tools a caller may use.

Recording agent runs

A shared trace ID connects the records of a run.

Spending limits for agent runs

Budget checks run before provider requests.

Pausing for human approval

The run waits for approval before the action executes.

Related projects

ClawKeeper

ClawKeeper checks proposed Claude agent actions against explicit rules before a model call. Autonomous spend defaults to zero, and the database enforces an append-only audit log.

TechTide Harness Kit

TechTide Harness Kit uses JSON Schema to declare skill permissions, risk tiers, network access, and approval requirements. Automated checks validate declarations and run prompt-injection fixtures.

CipherClaw

CipherClaw analyzes recorded agent activity without model calls. It builds causal graphs, scores possible causes, and flags failure patterns.

FintheFinder

FintheFinder pauses research runs for human approval and audits citations deterministically. Its repository includes contracts and test files.

I document the project here. TechTide AI publishes current service scope, prices, and terms.

Critical System Implementation at TechTide AI
Back to top
Talk to Alex
alexcinovoj.dev

© 2026 Alex Cinovoj · TechTide AI · Columbus, OH