Compare
CodeHerder vs the agent-orchestration field
A cell-by-cell look at CodeHerder against the agent-orchestration field: what's published, what's actually different, and what we couldn't verify.
Before the table
Three shapes of tool
Most of what gets called "AI coding agent" today is one of three fairly different shapes, and it's worth knowing which one you're looking at before comparing feature lists.
The first is an assistant that lives in one person's editor or terminal: you drive it, turn by turn, and it's good at that job. The second is a single autonomous agent that runs in a vendor's own cloud: you hand it a task and it works unattended, but it's still one agent, on infrastructure someone else owns and bills you for. The third is a herd: many agents, working many tasks at once, on machines you already own, staffed and governed by a control plane that treats the work itself, not the infrastructure, as the thing being managed.
CodeHerder is the third shape. The first is a great fit for one developer working alone; the second is a great fit for a single well-defined task you're happy to hand off entirely. Neither is wrong, they just answer a different question than "how do I run a team's worth of AI coding work, visibly and accountably, on hardware I already have." The table below compares CodeHerder against the other vendors building toward that third shape, plus a few adjacent single-agent products worth knowing about.
The capability matrix
20 capabilities, 15 platforms
Every cell traces to a source dated below. Read this before the table:
- ✅ published + sourced
- the vendor's own site/docs/announcement states this, with a source cited on the competitor's own page below.
- ⚠️ partial/qualified
- published, but narrower, differently-scoped, or hedged relative to CodeHerder's own version of the capability; see the footnote where one exists.
- "not published"
- no public evidence was found. This is not proof the vendor is missing the feature — it may exist and simply not be marketed. Never read a "not published" cell as a competitive claim of absence.
- unverified
- some signal exists (a transcript mention, a third-party comparison) but no primary source was located this pass.
"Not published" means no public evidence was found, not that the feature doesn't exist. A vendor may build and simply not market something. Treat every "not published" cell as absence of evidence, never as evidence of absence.
Verified as of 2026-08-08.
| Capability | CodeHerder | Orchestratia | Gas City/Gasworks | Warp Oz | Coder Agents | GitHub Agent HQ | Stoneforge | Factory | Devin | Codex app + Symphony | Claude Code Agent Teams | Augment Intent | OpenHands | Verdent | Cursor |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Concurrent agents | Published | Published | Published | Published | Published | Published | Published | Published | Published | Published | Published | Published | Published | Published | Published |
| Multi-vendor harnesses | Published | Published | Published | Published | Published | Published | Published | Not published (as of 2026-08-08) [1] | Not published (as of 2026-08-08) [1] | Partial [2] | Not published (as of 2026-08-08) [1] | Published | Published | Unverified | Not published (as of 2026-08-08) [1] |
| Customer-owned machines | Published | Published | Published | Published | Published | Partial [3] | Published | Published | Not published (as of 2026-08-08) | Partial [4] | Partial [4] | Partial [4] | Published | Unverified | Not published (as of 2026-08-08) |
| Multi-machine routing | Published | Published | Partial [5] | Published | Published | Partial [3] | Not published (as of 2026-08-08) | Published | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) | Published | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) |
| Automatic work assignment | Published | Published | Published | Published | Partial [6] | Partial [6] | Published | Published | Published | Published | Published | Published | Unverified | Published | Not published (as of 2026-08-08) |
| Capability-based routing | Published | Published | Partial (configurable) | Unverified | Unverified | Partial | Not published (as of 2026-08-08) | Published | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) | Partial (human-picked tier) | Partial | Not published (as of 2026-08-08) | Partial | Not published (as of 2026-08-08) |
| Atomic claiming | Published | Unverified | Partial (work queue) | Unverified | Unverified | Not published (as of 2026-08-08) | Published | Unverified | Unverified | Partial | Partial | Unverified | Unverified | Unverified | Not published (as of 2026-08-08) |
| Dependencies | Published | Published | Published | Partial | Not published (as of 2026-08-08) | Partial (issue-based) | Published | Partial | Partial | Not published (as of 2026-08-08) | Published | Partial | Not published (as of 2026-08-08) | Partial | Not published (as of 2026-08-08) |
| Custom workflows | Published | Partial (approval rules, not a stage schema) | Published (fully programmable) | Published | Partial (governance, not authoring) | Partial (Actions integration) | Partial (one fixed pipeline) | Partial | Not published (as of 2026-08-08) | Published (spec) | Not published (as of 2026-08-08) | Partial | Unverified | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) |
| Server-enforced gates | Published | Published | Published | Partial | Partial | Partial | Published | Unverified | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) | Published | Not published (as of 2026-08-08) | Partial | Not published (as of 2026-08-08) |
| Agent handoffs | Published | Published (contracts) | Published | Partial (context) | Not published (as of 2026-08-08) | Partial (PR/context) | Published | Partial | Published (coordinator) | Partial | Published (mailbox) | Published (spec-based) | Not published (as of 2026-08-08) | Unverified | Not published (as of 2026-08-08) |
| Persistent shared memory | Published | Not published (as of 2026-08-08) | Published (packs/state) | Published (preview) | Unverified | Partial (repo/AGENTS.md) | Partial (JSONL state) | Unverified | Partial (trajectory learning) | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) | Partial (context engine) | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) |
| Live terminal / view | Published | Published | Published | Partial (logs) | Partial | Partial (session logs) | Published | Unverified | Published (session link) | Partial (diff thread) | Partial (optional tmux) | Published | Unverified | Unverified | Unverified |
| Human takeover / steer | Published | Published | Published | Published | Unverified | Published | Partial | Unverified | Published | Partial | Published | Published | Unverified | Unverified | Unverified |
| Cost per task | Published | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) | Partial (credits) | Unverified | Partial (metrics dashboard) | Not published (as of 2026-08-08) | Unverified | Published (ACU) | Not published (as of 2026-08-08) | Partial (token usage) | Partial (credit pool) | Unverified | Unverified | Unverified |
| Cost per stage/model | Published | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) |
| Hard $ cap | Published | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) | Published (compute-budget control) | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) |
| Schedules | Published | Unverified | Published | Published | Not published (as of 2026-08-08) | Published (Actions) | Not published (as of 2026-08-08) | Unverified | Published | Published (Automations) | Not published (as of 2026-08-08) | Unverified | Unverified | Unverified | Unverified |
| Webhooks / API / CLI | Published | Published | Published | Published | Published | Published | Published | Unverified | Published | Published | Partial (CLI only) | Unverified | Published | Unverified | Unverified |
| Self-host full platform | Published (Enterprise) | Partial (Enterprise/custom) | Published (OSS) | Published (Enterprise) | Published | Not published (as of 2026-08-08) | Published (OSS) | Partial (Droid Computers only) | Not published (as of 2026-08-08) | Partial (spec only, not the app) | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) | Published | Not published (as of 2026-08-08) | Not published (as of 2026-08-08) |
- Factory (Droids), Devin, and Claude Code Agent Teams are each vertically integrated — the product's whole thesis is a single vendor's own agent(s), so "multi-vendor harness support" isn't just unpublished, it's outside the product's own stated shape. Still written as "not published" rather than a claim of proven absence, per this tree's own rule.
- The Codex *app* is Codex-only; the Symphony *spec* is model-agnostic by design and has community forks substituting Claude Code, but its reference implementation and default is Codex.
- GitHub Agent HQ orchestrates across github.com, VS Code, mobile, and CLI — powerful distribution, but not a "register the customer's own server" model the way Orchestratia's SSH-tunnel daemon or Factory's Droid Computers are.
- Codex app, Claude Code Agent Teams, and Augment Intent each run locally on the operator's own single machine (a real customer-owned-compute story for that one machine) but do not publish a *fleet* model spanning multiple customer-owned machines the way CodeHerder, Orchestratia, Gas City, Warp Oz, Coder Agents, Stoneforge, and Factory do.
- 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.
- Coder Agents and GitHub Agent HQ both provision/trigger work from many entry points (chat, API, CI, Slack, Jira for Coder; assign-across-repos for GitHub) but the assignment step itself reads as more human-directed than an automatic router picking the agent.
Notice a cell that's out of date? Tell us and we'll fix it.
Every platform on the table
Read the full page for any of them
Each page names the pages we checked and dates what we verified.
Direct shape match
Orchestratia
A multi-agent orchestration dashboard for coordinating Claude Code, Gemini CLI, and Codex across servers you already run.
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.
Warp Oz
Warp's cloud orchestration platform for coding agents: hundreds of parallel, sandboxed agents behind a multi-harness control plane.
Coder Agents
A fully self-hosted, model-agnostic agent orchestrator from Coder, built for platform teams that need governance over models, prompts, and skills.
GitHub Agent HQ
GitHub's multi-vendor agent orchestration layer, built into the repository you're probably already using.
Stoneforge
An open-source (Apache-2.0) agent orchestrator: a Director plans, Workers execute in isolated worktrees, and Stewards test, merge, or hand back for review.
Factory
An SDLC-wide agent platform for mid to large organizations, including Droid Computers: persistent, remotely orchestrated machines you can register yourself.
Vertically integrated platform
Devin
Cognition's autonomous coding agent, now able to manage a hierarchy of other Devins on their own VMs.
OpenAI Codex app + Symphony
OpenAI's own command center for Codex agents, plus Symphony, an open orchestration spec that turns an issue tracker into the control plane.
Claude Code Agent Teams
An experimental Anthropic feature: a Claude Code team lead spawns teammates with their own context windows, connected by a shared task list and a direct mailbox.
Augment Intent
A Mac desktop workspace where a Coordinator agent proposes a plan, Implementor agents execute it in isolated worktrees, and a Verifier checks the result before handing back for review.
OpenHands
A self-hosted, model-agnostic open-source background-agent framework, positioned as the open platform for cloud coding agents.
Verdent
A parallel-first, AI-native development environment that decomposes goals into small, independently owned work items from the very first task.
Cursor
Cursor's background agents, tied to its own harness and billing, now paired with Graphite's code review after Cursor's December 2025 acquisition.
Round up your herd.
Bring every human and every agent onto one table. Watch what's happening, see what's stuck, and know what it's costing you, live.
Already have a workspace? Sign in