Sequential Thinking (Reference) vs agent-harness
Sequential Thinking (Reference) is an MCP server published by modelcontextprotocol. Marketplaces we read that list it: 1 (Docker MCP Catalog). agent-harness is a plugin bundle for AI coding tools published by Alireza Rezvani. Marketplaces we read that list it: 1 (Community GitHub marketplaces). As of our marketplace pass on 2026-10-15. This pair was picked by a reader, not by our comparison rule, so search engines are asked not to index the page.
Where each is published
| Marketplace |  Sequential Thinking (Reference) | |
|---|---|---|
| Docker MCP Catalog | Listed Version: sha256:cd3174b2ecf3 Forks (GitHub): 11 790 Pulls (Docker Hub): 422 526 Stars (Docker Hub): 30 Stars (GitHub): 91 268 Not marked · First seen |
Not listed |
| Community marketplace alirezarezvani/claude-skills | Not listed | Listed Version: 1.0.0 Forks (GitHub): 3 963 Stars (GitHub): 28 221 Not marked · First seen |
Each figure is the marketplace's own and is never added up across marketplaces. A marketplace that publishes no figure shows none: no number is not zero.
The facts side by side
| Fact | Sequential Thinking (Reference) | agent-harness |
|---|---|---|
| Marketplaces that list it | 1 | 1 |
| Stars (GitHub) (differs) | 91 268 (better) | 28 221 (worse) |
| Pulls (Docker Hub) (differs) | 422 526 | No figure on Docker Hub |
| Forks (GitHub) (differs) | 11 790 (better) | 3 963 (worse) |
| Last release | Not published | Not published |
| Where it runs (differs) | On your machine (mcp/sequentialthinking) | Installs as a plugin bundle |
| Transport (differs) | stdio | Not stated |
| Tools it declares (differs) | 1 | Not published |
| Environment variables it asks for | Not published | Not published |
| Parts of the bundle (differs) | Not a bundle | 3 |
| Licence (differs) | NOASSERTION | MIT |
| Repository (differs) | github.com/modelcontextprotocol/servers | github.com/alirezarezvani/claude-skills |
| Changes in the last 7 days (differs) | 0 | 2 |
| First seen (differs) | 2026-10-11 | 2026-10-14 |
Lines where the two differ have a tinted background and bold values. Where the data itself says which value is better (more marketplaces, a larger figure of the same marketplace, a fresher release, a verified mark on a marketplace that lists both), the better one is also marked green with ▲ and the worse one red with ▼. Where it runs, the transport, the tools, the settings, hooks, the bundle's parts, the licence and the week's changes are marked only as different: more or fewer is not better in itself.
What each says it does
Sequential Thinking (Reference)
Dynamic and reflective problem-solving through thought sequences.
agent-harness
Turn any domain folder of skills into a bounded agentic loop: a manifest builder inventories a domain's skills/tools/checks, a goal compiler turns a goal into a verifiable task plan (refusing vague goals with forcing questions), and a JSON-backed loop controller drives execute->verify->close with retry caps, controller-run verification (no verification theater), human escalation on exhausted budgets, and a close gate that refuses while any task is unverified or unwaived. Ships 3 stdlib Python tools, 18 committed per-domain harness manifests + JSON schema, 3 references citing the 2024-2026 agent-harness canon (Anthropic long-running harnesses, verifier's law, SWE-agent, Ralph loop, Cognition), harness-runner agent + /cs:harness command. Use when an agent or subagent should pick up a goal, define its tasks, complete and verify them, and close the loop.
How to run each
Sequential Thinking (Reference) As a local MCP server: Docker image
claude mcp add --transport stdio sequentialthinking -- docker run -i --rm mcp/sequentialthinkingagent-harness As a Claude Code plugin
/plugin marketplace add alirezarezvani/claude-skills
/plugin install agent-harness@claude-code-skillsCommands built only from what the marketplaces declare; the full set for every client is on each card.
Tools each declares
The marketplaces of agent-harness publish no list of tools, so we do not compare tools.
Settings each asks for
Neither one's marketplaces publish the settings it asks for.
What it can do: Sequential Thinking (Reference) · What it can do: agent-harness · Compare two other plugins
Data comes from the Docker MCP Catalog (desktop.docker.com/mcp/catalog) and Docker Hub. Data comes from public marketplace files in GitHub repositories of their authors. Figures come from our own regular passes over the marketplaces, last read on . We publish no contact details for plugin publishers.