{"id":49381590,"url":"https://github.com/scottconverse/agentsuite","last_synced_at":"2026-04-28T06:00:31.463Z","repository":{"id":354037447,"uuid":"1221795966","full_name":"scottconverse/AgentSuite","owner":"scottconverse","description":"Seven role-specific reasoning agents that turn vague intent into precise operating artifacts","archived":false,"fork":false,"pushed_at":"2026-04-26T21:42:38.000Z","size":122,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-04-26T22:07:15.499Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":"https://scottconverse.github.io/AgentSuite/","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/scottconverse.png","metadata":{"files":{"readme":"README.md","changelog":"CHANGELOG.md","contributing":"CONTRIBUTING.md","funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-04-26T17:35:00.000Z","updated_at":"2026-04-26T21:42:42.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/scottconverse/AgentSuite","commit_stats":null,"previous_names":["scottconverse/agentsuite"],"tags_count":null,"template":false,"template_full_name":null,"purl":"pkg:github/scottconverse/AgentSuite","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/scottconverse%2FAgentSuite","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/scottconverse%2FAgentSuite/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/scottconverse%2FAgentSuite/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/scottconverse%2FAgentSuite/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/scottconverse","download_url":"https://codeload.github.com/scottconverse/AgentSuite/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/scottconverse%2FAgentSuite/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":32368534,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-04-27T20:07:02.737Z","status":"online","status_checked_at":"2026-04-28T02:00:07.250Z","response_time":56,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":[],"created_at":"2026-04-28T06:00:19.952Z","updated_at":"2026-04-28T06:00:31.449Z","avatar_url":"https://github.com/scottconverse.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# AgentSuite\n\n\u003e Seven role-specific reasoning agents that turn vague intent into precise operating artifacts.\n\u003e\n\u003e **v0.8.1** — Specification Kernel + Founder · Design · Product · Engineering · Marketing · Trust/Risk · CIO Agents\n\nAgentSuite is a Python package and MCP server. It exposes role-specific agents (Founder, Design, Product, Engineering, Marketing, Trust/Risk, and CIO shipped) that take loose human intent and produce structured, reusable artifacts: brand systems, brief libraries, voice guides, prompt templates, engineering specs, IT strategies, and more.\n\nThe agents are reasoning agents, not content generators. Output is a reusable system, not a one-off asset.\n\n## Why this exists\n\nModern AI tools generate fast. They do not generate *consistently*. Without a reusable brand system and brief library, every new asset re-introduces context, drifts on voice, and slowly turns into generic SaaS copy. AgentSuite codifies the operating system around generation: every agent persists its output to disk, runs through a six-stage pipeline (intake → extract → spec → execute → qa → approval), and promotes approved artifacts into a long-lived `_kernel/` that downstream agents consume.\n\n## Install\n\n```bash\n# Install with your preferred provider:\npip install \"agentsuite[anthropic] @ git+https://github.com/scottconverse/AgentSuite.git\"   # Anthropic Claude\npip install \"agentsuite[openai] @ git+https://github.com/scottconverse/AgentSuite.git\"      # OpenAI GPT\npip install \"agentsuite[gemini] @ git+https://github.com/scottconverse/AgentSuite.git\"      # Google Gemini\npip install \"agentsuite[ollama] @ git+https://github.com/scottconverse/AgentSuite.git\"      # Local Ollama daemon\n\n# Install everything:\npip install \"agentsuite[all] @ git+https://github.com/scottconverse/AgentSuite.git\"\n\n# or, no install (MCP only):\nuvx --from \"agentsuite[mcp] @ git+https://github.com/scottconverse/AgentSuite.git\" agentsuite-mcp\n```\n\n\u003e AgentSuite is distributed from GitHub only — there is no PyPI publication.\n\nRequirements: Python 3.11+. Either an API key (`ANTHROPIC_API_KEY`, `OPENAI_API_KEY`, or `GEMINI_API_KEY` / `GOOGLE_API_KEY`) OR a local Ollama daemon running at `localhost:11434`.\n\n## Quick start (CLI)\n\n```bash\nexport ANTHROPIC_API_KEY=sk-ant-...\nagentsuite founder run \\\n  --business-goal \"Launch PatentForgeLocal v1\" \\\n  --project-slug pfl \\\n  --inputs-dir ./examples/patentforgelocal\n```\n\nThe agent walks five stages, writes 26 artifacts under `.agentsuite/runs/\u003crun-id\u003e/`, and pauses at the approval gate. Review `brand-system.md` + `qa_report.md`, then:\n\n```bash\nagentsuite founder approve --run-id \u003crun-id\u003e --approver you --project-slug pfl\n```\n\nApproved artifacts get promoted to `.agentsuite/_kernel/pfl/` for downstream agents.\n\n## Quick start (Local LLM — Ollama)\n\nRun AgentSuite entirely offline against a local Gemma 4 model. Zero cost, no API keys required.\n\n```bash\n# Install Ollama (https://ollama.ai) if not already installed, then pull a Gemma 4 model:\nollama pull gemma4:e4b      # recommended: ~5 GB, balanced quality\n\n# Optional alternatives:\n#   ollama pull gemma4:e2b    # ~3 GB, laptop-class (faster, less capable)\n#   ollama pull gemma4:26b-moe # ~15 GB, high-end workstation (slowest, best quality)\n\n# Make sure the daemon is running:\nollama serve \u0026\n\n# Then run AgentSuite as usual — the resolver auto-detects the daemon when no API keys are set:\nagentsuite founder run \\\n  --business-goal \"Launch My Product v1\" \\\n  --project-slug my-product \\\n  --inputs-dir ./my-brand-inputs\n```\n\nYou can also pin Ollama explicitly via `AGENTSUITE_LLM_PROVIDER=ollama` regardless of which API keys you have set.\n\n## Quick start (MCP — Codex)\n\nAdd to `~/.codex/mcp.toml`:\n\n```toml\n[servers.agentsuite]\ncommand = \"uvx\"\nargs = [\"agentsuite-mcp\"]\n\n[servers.agentsuite.env]\nAGENTSUITE_ENABLED_AGENTS = \"founder,design,product,engineering,marketing,trust-risk,cio\"\n```\n\nRestart Codex. Tools `founder_run`, `founder_approve`, `founder_get_status`, `founder_list_runs`, `founder_resume`, plus the cross-agent `agentsuite_list_agents`, `agentsuite_kernel_artifacts`, `agentsuite_cost_report` are now callable.\n\n## Quick start (MCP — Claude Code / Cowork)\n\nAdd to project-root `.mcp.json`:\n\n```json\n{\n  \"mcpServers\": {\n    \"agentsuite\": {\n      \"command\": \"uvx\",\n      \"args\": [\"agentsuite-mcp\"],\n      \"env\": {\"AGENTSUITE_ENABLED_AGENTS\": \"founder,design,product,engineering,marketing,trust-risk,cio\"}\n    }\n  }\n}\n```\n\nRestart the harness.\n\n## What the agents produce\n\n### Founder agent — 26 artifacts per run:\n\n| Stage | Artifacts |\n|---|---|\n| 1 intake | `inputs_manifest.json` |\n| 2 extract | `extracted_context.json` |\n| 3 spec | `brand-system.md`, `founder-voice-guide.md`, `product-positioning.md`, `audience-map.md`, `claims-and-proof-library.md`, `visual-style-guide.md`, `campaign-production-workflow.md`, `asset-qa-checklist.md`, `reusable-prompt-library.md`, `consistency_report.json` |\n| 4 execute | `brief-template-library/` (11 brief templates) + `export-manifest-template.json` |\n| 5 qa | `qa_report.md`, `qa_scores.json` |\n| state | `_state.json`, `_meta.json` |\n\nOn `founder_approve`, the spec artifacts + brief-template-library are promoted to `.agentsuite/_kernel/\u003cproject_slug\u003e/` for use by downstream agents.\n\n### Design agent — see CHANGELOG for artifact list.\n\n### Product agent — see CHANGELOG for artifact list.\n\n### Engineering agent — 17 artifacts per run:\n\n| Stage | Artifacts |\n|---|---|\n| 3 spec | `architecture-decision-record.md`, `system-design.md`, `api-spec.md`, `data-model.md`, `security-review.md`, `deployment-plan.md`, `runbook.md`, `tech-debt-register.md`, `performance-requirements.md` |\n| 4 execute | `brief-template-library/` (8 brief templates: sprint ticket, code review checklist, incident report, capacity plan, on-call handoff, release checklist, postmortem, vendor evaluation) |\n| 5 qa | `qa_report.md`, `qa_scores.json` |\n| state | `_state.json`, `_meta.json` |\n\n### Trust/Risk agent (v0.6.0) — 17 artifacts per run:\n\nInput: `product_name`, `risk_domain`, `stakeholder_context` (plus optional `regulatory_context`, `threat_model_scope`, `compliance_frameworks`).\n\n| Stage | Artifacts |\n|---|---|\n| 3 spec | `threat-model.md`, `risk-register.md`, `control-framework.md`, `incident-response-plan.md`, `compliance-matrix.md`, `vendor-risk-assessment.md`, `security-policy.md`, `audit-readiness-report.md`, `residual-risk-acceptance.md`, `consistency_report.json` |\n| 4 execute | `brief-template-library/` (8 brief templates: breach-notification, executive-risk-summary, penetration-test-brief, remediation-tracker, risk-acceptance-form, security-awareness-brief, tabletop-exercise-scenario, vendor-security-questionnaire) + `export-manifest-template.json` |\n| 5 qa | `qa_report.md`, `qa_scores.json` |\n| state | `_state.json`, `_meta.json` |\n\nPrimary artifact: `threat-model.md`. On `trust_risk_approve`, all spec artifacts and brief templates are promoted to `.agentsuite/_kernel/\u003cproject_slug\u003e/`.\n\n### CIO agent (v0.7.0) — 17 artifacts per run:\n\nInput: `organization_name`, `strategic_priorities`, `it_maturity_level`.\n\n| Stage | Artifacts |\n|---|---|\n| 3 spec | `it-strategy.md`, `technology-roadmap.md`, `vendor-portfolio.md`, `digital-transformation-plan.md`, `it-governance-framework.md`, `enterprise-architecture.md`, `budget-allocation-model.md`, `workforce-development-plan.md`, `it-risk-appetite-statement.md`, `consistency_report.json` |\n| 4 execute | `brief-template-library/` (8 brief templates: board-technology-briefing, it-steering-committee-agenda, vendor-review-summary, project-portfolio-status, digital-initiative-proposal, it-investment-case, technology-modernization-pitch, quarterly-it-review) + `export-manifest-template.json` |\n| 5 qa | `qa_report.md`, `qa_scores.json` |\n| state | `_state.json`, `_meta.json` |\n\nPrimary artifact: `it-strategy.md`. On `cio_approve`, all spec artifacts and brief templates are promoted to `.agentsuite/_kernel/\u003cproject_slug\u003e/`.\n\n## Quick start (Python SDK)\n\nUse AgentSuite programmatically — no CLI required:\n\n```python\nfrom pathlib import Path\nfrom agentsuite import FounderAgent, resolve_provider\nfrom agentsuite.agents.founder.input_schema import FounderAgentInput\nfrom agentsuite.kernel.schema import Constraints\n\nagent = FounderAgent(output_root=Path(\".agentsuite\"), llm=resolve_provider())\nstate = agent.run(\n    request=FounderAgentInput(\n        agent_name=\"founder\",\n        role_domain=\"product\",\n        user_request=\"My startup builds X for Y\",\n        business_goal=\"Validate and launch\",\n        project_slug=\"my-startup\",\n        constraints=Constraints(),\n    ),\n    run_id=\"my-run\",\n)\nprint(state.stage)   # \"approval\"\n```\n\n`resolve_provider()` auto-detects your API key from environment variables. See [Configuration](#configuration) for the full list of env vars.\n\n## Configuration\n\n| Env var | Default | Purpose |\n|---|---|---|\n| `AGENTSUITE_ENABLED_AGENTS` | `founder` | Comma-separated agent names to expose (e.g. `founder,design,product,engineering`) |\n| `AGENTSUITE_OUTPUT_DIR` | `.agentsuite` | Where artifacts are written |\n| `AGENTSUITE_LLM_PROVIDER` | (auto-detect) | Force `anthropic`, `openai`, `gemini`, or `ollama` |\n| `AGENTSUITE_COST_CAP_USD` | `5.0` | Hard kill cap per run |\n| `AGENTSUITE_EXPOSE_STAGES` | (off) | Set `true` to expose `founder_intake`/`extract`/`spec`/`execute`/`qa` as MCP tools |\n\n## Architecture\n\nAgentSuite runs every request through a five-stage pipeline:\n\n```mermaid\nflowchart LR\n    A[Intake] --\u003e B[Extract]\n    B --\u003e C[Spec]\n    C --\u003e D[Execute]\n    D --\u003e E[QA]\n    E --\u003e|score ≥ 7.0| F[✅ Approved]\n    E --\u003e|score \u003c 7.0| G[🔄 Approval Gate]\n    G --\u003e|human approves| F\n```\n\nEach stage is handled by the same kernel regardless of which agent is running. The QA stage scores output against a 9-dimension rubric; runs scoring below 7.0 enter an approval gate before artifacts are written.\n\n```\n                    ┌───────────────┐\n                    │  Harness      │  (Codex / Claude Code / Cowork)\n                    └─────┬─────────┘\n                          │ stdio MCP\n                    ┌─────▼─────────┐\n                    │ agentsuite-mcp│\n                    └─────┬─────────┘\n                          │\n              ┌───────────┴───────────┐\n              │                       │\n        ┌─────▼──────┐         ┌──────▼──────┐\n        │  Kernel    │◄────────│ FounderAgent│\n        │            │         │             │\n        │ schema/    │         │ stages/     │\n        │ qa/        │         │ rubric.py   │\n        │ cost/      │         │ templates/  │\n        │ artifacts/ │         │ prompts/    │\n        │ approval/  │         └─────────────┘\n        │ base_agent │\n        └────────────┘\n```\n\nFull architecture diagram with all agents: see `docs/README-FULL.pdf`.\n\n## Status\n\n**Shipped:**\n- v0.1.0 — Specification Kernel + Founder Agent\n- v0.2.0 — Design Agent (brief generation, brand QA scoring)\n- v0.3.0 — Product Agent (PM intent → UI spec → coding handoff)\n- v0.4.0 — Engineering Agent (architecture decisions, system design, API specs, security review, deployment, runbook, tech-debt register, performance requirements)\n- v0.5.0 — Marketing Agent (campaign brief, audience profile, messaging framework, content calendar, channel strategy, SEO keyword plan, competitive positioning, launch plan, measurement framework)\n- v0.6.0 — Trust/Risk Agent (threat model, risk register, control framework, incident response plan, compliance matrix, vendor risk assessment, security policy, audit readiness report, residual risk acceptance)\n- v0.7.0 — CIO Agent (IT strategy, technology roadmap, vendor portfolio, digital transformation plan, IT governance framework, enterprise architecture, budget allocation model, workforce development plan, IT risk appetite statement)\n\n**Roadmap:**\n- v0.8.0 — next agent\n\n## Documentation\n\n- [USER-MANUAL.md](docs/USER-MANUAL.md) — plain-language walkthrough\n- [README-FULL.pdf](docs/README-FULL.pdf) — full reference with architecture diagrams\n- [CONTRIBUTING.md](CONTRIBUTING.md) — dev setup + agent-implementation guide\n- [CHANGELOG.md](CHANGELOG.md) — release notes\n\n## License\n\nMIT — see [LICENSE](LICENSE).","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fscottconverse%2Fagentsuite","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fscottconverse%2Fagentsuite","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fscottconverse%2Fagentsuite/lists"}