Compare · Tier A
CodeHerder vs Gas City / Gasworks
Steve Yegge's programmable agent-orchestration platform: a durable, Git-backed issue tracker plus TOML formulas that compile into multi-stage agent workflows.
- 8 August 2026
- Last checked
- 20
- Capabilities scored, side by side
Verified against the vendor's own published materials.
What Gas City / Gasworks actually ships
- Beads, a Git-backed durable work-unit tracker where every item, including mail and agent identity, is a Bead
- Programmable formulas: TOML workflow definitions compiled into sequential-stage, dependency-graph, fan-out, loop, and error-policy pipelines
- A review-and-gap-check pass after each run, comparing the result against the plan and filling what's missing
- Gasworks, a managed multi-operator layer with shared workflows and live visibility across operators, SaaS or self-hosted
Where it's strong
- Open source (MIT) and self-hostable, with a real, shipped fan-out-and-review pattern
- Arguably the more programmable orchestration engine of the two: a team that wants to author its own workflow graph from scratch has real room to do that here
How CodeHerder differs
- CodeHerder ships as a configured product; Gas City hands you the primitives to build your own factory
- CodeHerder pairs its workflow engine with cost and budget attribution, a device herd spanning multiple machines, and organizational (initiative → epic → story) structure, none of which Gas City publishes
Gas City is 'build your own software factory.' CodeHerder is 'here is your software factory.'
Capability snapshot
CodeHerder vs Gas City / Gasworks, capability by capability
Every row traces to the same source as the full matrix, where the full legend lives. In short: "not published" means no public evidence was found, not that the feature is absent. Verified as of 2026-08-08.
| Capability | CodeHerder | Gas City / Gasworks |
|---|---|---|
| Concurrent agents | Published | Published |
| Multi-vendor harnesses | Published | Published |
| Customer-owned machines | Published | Published |
| Multi-machine routing | Published | Partial [5] |
| Automatic work assignment | Published | Published |
| Capability-based routing | Published | Partial (configurable) |
| Atomic claiming | Published | Partial (work queue) |
| Dependencies | Published | Published |
| Custom workflows | Published | Published (fully programmable) |
| Server-enforced gates | Published | Published |
| Agent handoffs | Published | Published |
| Persistent shared memory | Published | Published (packs/state) |
| Live terminal / view | Published | Published |
| Human takeover / steer | Published | Published |
| Cost per task | Published | Not published (as of 2026-08-08) |
| Cost per stage/model | Published | Not published (as of 2026-08-08) |
| Hard $ cap | Published | Not published (as of 2026-08-08) |
| Schedules | Published | Published |
| Webhooks / API / CLI | Published | Published |
| Self-host full platform | Published (Enterprise) | Published (OSS) |
- Gas City is self-hostable and OSS, which implies multi-machine capability, but no source found describes an explicit cross-machine routing mechanism the way Orchestratia's multi-server dashboard or Warp Oz's fleet model do.
Sources
Verified against Gas City / Gasworks's own site as of 2026-08-08.
- Gas City's documentation site
- Gas City's Gasworks page
Round up your herd.
Bring every human and every agent onto one table. Watch the work move. Costs update as it happens.