CodeZaiku vs Stepgate
A side-by-side look at the CodeZaiku and Stepgate MCP servers: transport, auth, tools, quality signals, and the config to install either one in Claude, Cursor, Windsurf or VS Code.
Verdict: choose CodeZaiku for coding agents work as a local process; choose Stepgate for coding agents as a local process. You can also run both side by side in one mcpServers block.
Which MCP server should you choose?
Choose CodeZaiku when
- You need dedicated capabilities in the Coding Agents domain.
- You prefer a local stdio subprocess transport.
- You have the keys it needs:
CODEZAIKU_PREFIX, CODEZAIKU_VERSION, CODEZAIKU_RUNTIME. - You want tools like
code, fix, explore_and_fix.
Choose Stepgate when
- You need dedicated capabilities in the Coding Agents domain.
- You prefer a local stdio subprocess transport.
- You have the keys it needs:
TAVILY_API_KEY. - You want tools like
stepgate_call, stepgate_submit, stepgate_guide.
Feature & specification comparison
Rows where the two servers differ are marked.
| Spec | ||
|---|---|---|
| Category | Coding Agents | Coding Agents |
| Quality signal | 63/100 · Good | 63/100 · Good |
| Install command | Confirmed | Confirmed |
| Transport | Local subprocess (stdio) | Local subprocess (stdio) |
| Auth | Not declared | Not declared |
| Pricing | Not declared | Not declared |
| Required env vars | CODEZAIKU_PREFIXCODEZAIKU_VERSIONCODEZAIKU_RUNTIME | TAVILY_API_KEY |
| Verified owner | Community listing | Community listing |
| GitHub stars | 1 | 13 |
| Directory activity | 1 views 0 installs 0 upvotes | 1 views 0 installs 0 upvotes |
| Tools | 11 | 8 |
| Full listing | View CodeZaiku | View Stepgate |
Tools & capabilities
CodeZaiku tools 11
codeRun CodeZaiku's coding familiar on an existing project: it reads the codebase, edits files, runs the toolchain, and verifies via the project's own tests. Use for bug-fix, add-feature, refactor, test-coverage, incident, and API-migration tasks. Blocks until the loop finishes.fixRun CodeZaiku's autonomous-SRE operator on a stack or host: senses health, localizes the root cause, applies a fix (validated card or read-only-recon-derived), verifies, rolls back on failure, and harm-checks. Bounded by blast-radius and an authority ladder (observe<localize<propose<guarded<unattended). Blocks until it concludes.explore_and_fixExplore an entire machine and fix what's broken ? boundary = THAT machine. Enumerates all docker compose stacks, failed systemd units, and host disk/memory pressure; runs the SRE operator on each unhealthy compose stack (blast-radius + authority-ladder bounded); surfaces systemd/host issues for attention. Use to autonomously triage a whole box, not one known stack.reviewREAD-ONLY code review: navigate a project (or a git diff) and return a structured findings list ([high|med|low] file:line ? problem ? suggested fix) covering correctness, security, and maintainability. Modifies nothing.researchResearch a question on the WEB (searches, fetches and reads real sources, then synthesizes a cited answer). mode=broad surveys the landscape across many sources; mode=depth reads the best sources thoroughly to answer one narrow question. Read-only.research_memoryInspect the RESEARCH MEMORY POOL ? findings accumulated from earlier research runs. Omit query for a summary of what has been researched; pass a query to recall relevant prior findings (so you can build on them instead of re-researching).investigateREAD-ONLY incident diagnosis / RCA on a stack or host: senses and localizes the root cause but does NOT remediate. The safe 'just tell me what's wrong' entry point.secureSECURITY review of a machine or stack: static posture (privileged containers, docker.sock mounts, datastores published on 0.0.0.0, weak/default credentials, image CVEs) PLUS runtime intrusion detections from the syscall detector, and for a live detection a read-only investigation proposing ONE surgical containment command. REPORTS ONLY ? it never applies a containment action, because a wrong containment is a self-inflicted outage.record_conventionRecord a CONVENTION for a project ? a norm the familiar should follow HERE (typically from a bondholder accept/correct on its work). Builds the project's coding DNA, which is loaded into context on future `code` runs for that project.show_conventionsShow a project's accumulated conventions (its coding DNA).job_statusPoll an async tool job submitted with async:true. Returns 'running', or the final result when the job is done/failed.
Stepgate tools 8
stepgate_callCalls one of the current step's operations for a Stepgate run and returns its result. Stepgate makes the request and adds any credentials.stepgate_submitSubmits the current step's output for a Stepgate run. Returns the next step, the gate failures to fix, or the finished run's outputs.stepgate_guideReturns how to write a stepfile: the authoring workflow, the full format reference and the JSON Schema. Call it before writing or editing a stepfile.stepgate_examplesLists the catalog's stepfiles, or with a name returns that stepfile's YAML and README, to copy a working pattern from.stepgate_validateChecks a draft stepfile against the schema and every load-time rule. Returns each issue with its path, or the stepfile's identity and whether stepgate_try accepts it.stepgate_inspect_apiFetches a public OpenAPI document or MCP server without credentials and reports what a stepfile needs: the document's sha256, servers, security schemes and operationIds, or an MCP server's tools and their schema_sha256.stepgate_outlineTurns a SKILL.md or markdown SOP into a skeleton stepfile: one step per numbered item or second-level heading, each rule it states marked as a gate to write, and TODO markers where only an author can decide. stepgate_validate lists the markers left.stepgate_tryStarts a run of a draft stepfile from its text, without adding it to the server. Drive it with stepgate_call and stepgate_submit. Drafts may call only public https URLs, and use only the credentials and settings the operator granted to drafts.
Ready-to-paste client configs
Paste either block (or both) into your client config, e.g. claude_desktop_config.json or ~/.cursor/mcp.json. We only show a config when the install command is confirmed.
CodeZaiku
{
"mcpServers": {
"codezaiku": {
"command": "npx",
"args": [
"-y",
"@wyrdsekai/codezaiku-mcp"
],
"env": {
"CODEZAIKU_PREFIX": "YOUR_CODEZAIKU_PREFIX_HERE",
"CODEZAIKU_VERSION": "YOUR_CODEZAIKU_VERSION_HERE",
"CODEZAIKU_RUNTIME": "YOUR_CODEZAIKU_RUNTIME_HERE"
}
}
}
}
Stepgate
{
"mcpServers": {
"stepgate": {
"command": "npx",
"args": [
"-y",
"stepgate"
],
"env": {
"TAVILY_API_KEY": "YOUR_TAVILY_API_KEY_HERE"
}
}
}
}
Frequently asked questions
CodeZaiku is categorized under Coding Agents and uses a local stdio subprocess. Stepgate belongs to Coding Agents and uses a local stdio subprocess. Pick CodeZaiku for coding agents work and Stepgate for coding agents.
Copy the mcpServers block from the config section on this page into your client's config (for example claude_desktop_config.json or ~/.cursor/mcp.json), then fully restart the client.
CodeZaiku has no declared pricing — check the project's own site. Stepgate has no declared pricing — check the project's own site. If a server needs environment variables such as API keys, define them under the env key in your client's MCP config.
Yes. MCP clients support multiple servers: add codezaiku and stepgate as separate keys in the same mcpServers object, and your assistant will call tools from either one.
On AllMCPs, CodeZaiku has a quality signal of 63/100 (Good) with 1 views, 0 installs and 1 GitHub stars. Stepgate has 63/100 (Good) with 1 views, 0 installs and 13 GitHub stars.