coeus
OfficialServer Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| COEUS_AGY_COMMAND | No | OAuth CLI executable. Default: agy | agy |
| COEUS_TOOL_PROFILE | No | Tool profile label returned by capabilities. Default: standard | standard |
| COEUS_AGY_STATE_DIR | No | Ignored, non-secret Agy runtime profile directory. Default: .coeus/state | .coeus/state |
| COEUS_CODEX_COMMAND | No | OAuth CLI executable. Default: codex | codex |
| COEUS_CLAUDE_COMMAND | No | OAuth CLI executable. Default: claude | claude |
| COEUS_PROVIDER_ORDER | No | Optional comma-separated provider preference order; capability and readiness checks still apply. Default: blank | |
| COEUS_ALIBABA_API_KEY | No | Direct Alibaba Cloud Model Studio credential. Default: empty | |
| COEUS_DEEPSEEK_API_KEY | No | Direct DeepSeek credential. Default: empty | |
| COEUS_DEFAULT_PROVIDER | No | Automatic routing or one canonical provider ID. Default: auto | auto |
| COEUS_ENABLE_WEB_TOOLS | No | Web capability pack flag. Default: true | true |
| COEUS_PROVIDER_TIMEOUT | No | Per-provider execution budget in seconds. Default: 120 | 120 |
| COEUS_AGY_DEFAULT_MODEL | No | Agy starting model. Default: gemini-3.7-flash-medium | gemini-3.7-flash-medium |
| COEUS_ENABLE_PIXEL_TOOLS | No | Pixel capability pack flag. Default: true | true |
| COEUS_OPENROUTER_API_KEY | No | Direct OpenRouter credential. Default: empty | |
| COEUS_CODEX_DEFAULT_MODEL | No | Codex starting model. Default: gpt-5.6-terra | gpt-5.6-terra |
| COEUS_DEEPSEEK_FAST_MODEL | No | DeepSeek fast model. Default: deepseek-v4-flash | deepseek-v4-flash |
| COEUS_ENABLE_DESIGN_TOOLS | No | Design capability pack flag. Default: true | true |
| COEUS_CLAUDE_DEFAULT_MODEL | No | Claude starting model. Default: claude-opus-5 | claude-opus-5 |
| COEUS_PROVIDER_CATALOG_TTL | No | Model catalog cache lifetime in seconds. Default: 60 | 60 |
| COEUS_ALIBABA_DEFAULT_MODEL | No | Alibaba starting model. Default: qwen3.8-max | qwen3.8-max |
| COEUS_DEEPSEEK_DEFAULT_MODEL | No | DeepSeek starting model. Default: deepseek-v4-pro | deepseek-v4-pro |
| COEUS_OPENROUTER_DEFAULT_MODEL | No | OpenRouter starting model. Default: openai/gpt-4.1-nano | openai/gpt-4.1-nano |
| COEUS_ENABLE_EXPERIMENTAL_TOOLS | No | Experimental workflow flag. Default: false | false |
| COEUS_PROVIDER_MAX_OUTPUT_BYTES | No | Bounded CLI output capture. Default: 1048576 | 1048576 |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": false
} |
| experimental | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| coeus_healthcheckB | Validate Coeus entry point, providers, keys, CLI paths, and dependencies |
| coeus_capabilitiesA | Return a truthful live inventory of tools, providers, capability packs, and experimental flags |
| coeus_list_modelsB | Return the authoritative model inventory for each configured provider |
| coeus_runtime_diagnosticsC | Return secret-safe runtime, CLI, git, catalog, and provider readiness diagnostics |
| coeus_statsC | Return consolidated Coeus server, cache, and rate-limit statistics |
| coeus_routeB | Choose provider, model, and workflow based on task, budget, speed, and vision needs |
| coeus_askC | Make a provider-neutral prompt call through the routed model provider |
| coeus_project_intakeC | Map a repository's stack, risks, key files, and suggested Coeus workflows |
| coeus_review_bundleB | Alias of coeus_review: bundle diff/code/context review and return structured findings. preview_only returns prompts and route without calling the provider. |
| coeus_counterfactual_reviewC | Review a change against alternate futures to expose hidden state, dependency, and release risks |
| coeus_debug_playbookB | Create debugging hypotheses, experiments, and logging or instrumentation plan |
| coeus_test_strategyC | Create a test plan before generating test code |
| coeus_docs_drift_checkC | Compare README/docs/examples against the live registry |
| coeus_release_readinessC | Return a ship/no-ship release readiness checklist |
| coeus_multi_model_verifyB | Run real verification across configured providers and summarize consensus limits |
| coeus_wildcardC | Generate three strange solution options and synthesize one strongman idea |
| coeus_design_review_bundleC | Review screenshot, HTML, and CSS artifacts with concrete implementation fixes |
| coeus_versionA | Return the installed Coeus package version without invoking a provider |
| coeus_analyzeC | Umbrella for summarize, codebase map, and explain; mode is summarize, map, or explain |
| coeus_summarizeB | Summarize content or file references; when files is provided it is used instead of content |
| coeus_eval_planC | Evaluate an implementation plan for feasibility, risk, and requirement fit |
| coeus_review_codeC | Review a code blob for quality, security, performance, and requested severity/output constraints |
| coeus_verify_solutionC | Verify a solution against requirements, test criteria, and context |
| coeus_thinkdeepC | Run a structured deep investigation of a bug, decision, or risk |
| coeus_project_task_planC | Convert project goals and tasks into an ordered implementation plan |
| coeus_second_opinionB | Get a rigorous second opinion on an approach or technical decision |
| coeus_workflow_automation_planC | Plan safe automation for a repetitive workflow |
| coeus_frontend_design_briefC | Create a frontend design brief; pipeline step 1 before taste direction, components, and assets |
| coeus_frontend_taste_directionA | Define non-generic frontend taste direction; pipeline step 2 before component creation |
| coeus_css_component_creatorC | Create stylesheet-backed component code from an approved design direction; pipeline step 3 |
| coeus_ui_asset_sourcesB | Find licensed UI asset sources and manifests; pipeline step 4 before static asset planning |
| coeus_static_asset_planB | Plan safe static asset integration from a sourced asset manifest; pipeline step 5 |
| coeus_critical_debateA | Execute a Red Team vs Blue Team adversarial debate on a plan, design, or code snippet to identify critical flaws and mitigations |
| coeus_analyze_codebaseC | Ask a large-context question about a directory or @file set with ignore guidance |
| coeus_codebase_mapC | Produce a large-context architecture map for a directory or @file set |
| coeus_design_reviewB | Review a screenshot, rendered HTML/CSS, or design artifact against frontend taste criteria |
| coeus_start_conversationC | Start an in-memory conversation session |
| coeus_continue_conversationA | Continue an existing in-memory conversation session |
| coeus_list_conversationsB | List active in-memory conversations |
| coeus_clear_conversationC | Clear an in-memory conversation session |
| coeus_ai_collaborationC | Prepare a multi-round provider collaboration workflow |
| coeus_extract_structuredC | Extract structured data from content using a JSON schema |
| coeus_git_diff_reviewA | Alias of coeus_review: review an explicit git diff and return structured findings (status/ship/findings). |
| coeus_reviewB | Structured finder-then-verify code review returning findings JSON. effort is quick, standard, or thorough. |
| coeus_content_comparisonC | Compare multiple sources semantically, structurally, factually, or as code |
| coeus_review_prB | Alias of coeus_review: gather a PR or base..head diff then return structured findings. |
| coeus_web_searchC | Requires a configured retrieval backend and returns an error without one; does not search the web by prompting a chat model |
| coeus_web_fetchB | Fetch a URL with SSRF protections and analyze fetched bytes; HTTP failures return error and do not invent page content |
| coeus_image_generateC | Generate image bytes only when a capable provider is configured; returns an error if the routed model cannot emit an image |
| coeus_generate_testsC | Generate test files or test code for supplied source |
| coeus_refactor_codeC | Suggest or generate refactors for supplied source code |
| coeus_explain_codeC | Explain supplied source code with optional symbol, line, and topic focus |
| coeus_generate_docsC | Generate docs, README content, API docs, or docstrings for supplied code |
| coeus_generate_commitC | Generate commit message candidates from a diff |
| coeus_pixel_art_generateC | Generate pixel-art sprite plans and renderable pixel grids |
| coeus_pixel_art_setB | Generate a matching UI icon family from names and a shared prompt |
| coeus_pixel_art_editC | Edit a pixel grid under a style contract (recolor or regenerate_slot) |
| coeus_pixel_art_findC | Find pixel-art references or sprites from allowed source families |
| coeus_pixel_paletteA | List or inspect bundled pixel-art palettes |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 59 tools
Multiple tools overlap heavily: `coeus_review_bundle`, `coeus_git_diff_review`, and `coeus_review_pr` are explicit aliases of `coeus_review`, while `coeus_review_code` and the design-review pair blur boundaries further. The analyze/summarize/explain/codebase-map cluster and the diagnostics cluster (stats, healthcheck, runtime_diagnostics, capabilities) also make tool selection ambiguous. An agent would frequently need extra context to avoid picking a near-duplicate.
All tools share the `coeus_` prefix and snake_case, which keeps the set readable, but verb/noun ordering is inconsistent: `generate_commit` contrasts with `pixel_art_generate`, and review aliases use `review_pr`, `git_diff_review`, and `review_bundle` interchangeably. Noun-only names such as `stats`, `version`, and `healthcheck` add to the mixed pattern.
At 59 tools, this server far exceeds the 25+ 'too many' threshold, and much of the bulk comes from aliases and specialty pipelines like frontend design and pixel art rather than core capabilities. The core ask/model/route functionality needs only a handful of tools, so the set should be consolidated significantly.
Across its broad intended scope, the server covers many workflows: ask/route, code review, tests, refactoring, docs, planning, web fetch/search, conversations, and pixel/frontend pipelines. However, provider/model configuration operations are missing (only list and diagnostics exist), several apparent duplicates are aliases rather than new capabilities, and some workflows depend on external backends. These gaps are work-aroundable but noticeable.