aiSlang
Agents Catalog

Real agents, declared end-to-end.

Every entry is a self-contained aiSlang project — capability-based model selection, prompts, agents, deterministic compute skills, RAG knowledge, MCP tools, evals and a deploy target — reconstructed straight from the .ais source in the examples tree.

16projects
26agents
7multi-agent
9compute skills

orchestrator-agent

multi-module · v0.1.0
multi-agentRAG knowledgeskillspipelinereflector loopDocker Composeevals

v0.3 Week 20 Mon — Orchestrator agent (multi-module project).

4 agents1 skill5 pipelines4 models4 evals
Details

Agents

  • architect [architect] — model architect_brain · prompt architect_sys · RAG architect_kb
  • code_reviewer [code_reviewer] — model reviewer_brain · prompt reviewer_sys · RAG reviewer_kb
  • coder [coder] — model coder_brain · prompt coder_sys · RAG coder_kb
  • tester [tester] — model tester_brain · prompt tester_sys · RAG tester_kb

Skills

  • run_tests — cargo test --workspace --no-fail-fast → string

Composition

  • review_once: code_reviewer
  • from_locked_design: coder
  • full_design: use architect.architect → use coder.coder → use code_reviewer.code_reviewer → use tester.tester
  • coder_review_cycle: use coder.coder → use code_reviewer.code_reviewer
  • test_once: tester
  • review_cycle (loop): body coder_review_cycle, max 3

Models

  • architect_brain (chat) — prefer anthropic.claude_haiku_4_5 · fallback ollama.qwen3_8b
  • reviewer_brain (chat, reasoning) — prefer anthropic.claude_haiku_4_5 · fallback ollama.qwen3_8b
  • coder_brain (chat) — prefer anthropic.claude_haiku_4_5 · fallback ollama.qwen2_5_coder_7b
  • tester_brain (chat, reasoning) — prefer anthropic.claude_haiku_4_5 · fallback ollama.qwen3_8b

RAG knowledge

  • Vector store(s): architect_kb, reviewer_kb, coder_kb, tester_kb
View source on GitHub

multi-module-demo

multi-module · v0.1.0
multi-agentMCP tools

v0.3 Week 16 Mon scaffold — see the sibling `aislang.toml`.

2 agents1 model
Details

Agents

  • architect [architect] — model thinker · prompt system
  • coder [coder] — model thinker · prompt system · tools docs

Models

  • thinker (chat) — prefer anthropic.claude_haiku_4_5 · fallback ollama.llama_3_1_8b

RAG knowledge

  • Corpus: README.md, company-info.md
View source on GitHub

patterns-router-demo

single-file
multi-agentMCP toolspipeline

v0.3 Week 17 Fri — worked router-pattern demo.

4 agents3 pipelines1 model
Details

Agents

  • classifier — model thinker · prompt classifier_sys
  • billing — model thinker · prompt billing_sys
  • technical — model thinker · prompt technical_sys
  • general — model thinker · prompt general_sys

Composition

  • handle_billing: billing
  • handle_technical: technical
  • handle_general: general

Models

  • thinker (chat) — prefer anthropic.claude_haiku_4_5 · fallback ollama.llama_3_1_8b

RAG knowledge

  • Corpus: README.md, company-info.md
View source on GitHub

patterns-orchestrator-demo

single-file
multi-agentMCP toolspipelineDocker Compose

v0.3 Week 17 Thu — worked orchestrator-pattern demo.

2 agents1 pipeline1 model
Details

Agents

  • researcher — model thinker · prompt researcher_sys · tools docs
  • summarizer — model thinker · prompt summarizer_sys

Composition

  • research_then_summarize: researcher → summarizer

Models

  • thinker (chat) — prefer anthropic.claude_haiku_4_5 · fallback ollama.llama_3_1_8b

RAG knowledge

  • Corpus: README.md, company-info.md
View source on GitHub

portfolio-pnl

single-file
multi-agentRAG knowledgeskillspipelineDocker Composeevals

A local-first, multi-agent portfolio analysis app built on [aiSlang](../../README.md). It helps one person understand a personal investment portfolio and make deliberate re-balancing decisions — and it runs **entirely on your own machine**.

2 agents3 skills1 pipeline1 model2 evals
Details

Agents

  • strategic — model reasoner · prompt strategic_sys · RAG strategy_kb
  • orchestrator — model reasoner · prompt orchestrator_sys · RAG strategy_kb

Skills

  • fetch_prices — python3 ../../../scripts/fetch.py → json
  • analyze — python3 ../../../scripts/analyze.py --persist → json
  • visualize — python3 ../../../scripts/render_dashboard.py → string

Composition

  • advise: strategic → orchestrator

Models

  • reasoner (chat, reasoning) — prefer anthropic.claude_haiku_4_5 · fallback ollama.qwen3_8b

RAG knowledge

  • Vector store(s): strategy_kb
  • Corpus: 01_antifragility.md, 02_barbell.md, 03_fragility_detection.md, 04_rebalancing.md, 05_reading_the_metrics.md
View source on GitHub

qwen-smoke

single-file
multi-agentDocker Compose

v0.3 Week 18 Fri — worked Qwen-on-Ollama smoke demo.

2 agents2 models
Details

Agents

  • thinker — model qwen3 · prompt qwen3_sys
  • coder — model qwen2_5_coder · prompt coder_sys

Models

  • qwen3 (chat) — prefer ollama.qwen3_8b
  • qwen2_5_coder (chat) — prefer ollama.qwen2_5_coder_7b
View source on GitHub

template-demo

single-file
multi-agentDocker Compose

v0.3 Week 18 Tue — worked dynamic-prompts demo.

2 agents1 model
Details

Agents

  • static_explainer — model thinker · prompt static_sys
  • dynamic_explainer — model thinker · prompt dynamic_sys

Models

  • thinker (chat) — prefer anthropic.claude_haiku_4_5 · fallback ollama.llama_3_1_8b
View source on GitHub

comment-clamper-agent

single-file
pipelineDocker Compose

v0.4 Week 26 Mon — comment-clamper dogfood agent.

1 agent1 pipeline1 model
Details

Agents

  • comment_clamper — model clamp_brain · prompt clamp_sys

Composition

  • clamp_once: comment_clamper

Models

  • clamp_brain (chat) — prefer anthropic.claude_haiku_4_5 · fallback ollama.qwen3_8b
View source on GitHub

db-agent

single-file
RAG knowledgepipelineDocker Composeevals

v0.3 Week 20 Wed — DB agent.

1 agent1 pipeline1 model1 eval
Details

Agents

  • db — model db_brain · prompt db_sys · RAG db_kb

Composition

  • brief_once: db

Models

  • db_brain (chat, reasoning) — prefer anthropic.claude_haiku_4_5 · fallback ollama.qwen3_8b

RAG knowledge

  • Vector store(s): db_kb
  • Corpus: 01_current_schemas.md, 02_proposed_sqlite_schema.md, 03_reversibility_contract.md
View source on GitHub

e2e-stack

single-file
Kubernetes

examples/e2e-stack/main.ais — self-contained K8s stack for the managed-cloud multi-tenant e2e (deploy/eks/40-e2e-two-tenants.sh).

1 agent1 model
Details

Agents

  • support — model fast · prompt system

Models

  • fast (chat) — prefer anthropic.claude_haiku_4_5, openai.gpt_4o_mini · fallback ollama.llama_3_1_8b
View source on GitHub

free-local

single-file
Docker Compose

examples/free-local.ais — the "no caps, free local model" demo.

1 agent1 model
Details

Agents

  • assistant — model local · prompt system

Models

  • local (chat) — prefer ollama.llama_3_1_8b

RAG knowledge

  • Corpus: README.md, company-info.md
View source on GitHub

rag-demo

single-file
RAG knowledgeDocker Compose

v0.3 Week 19 Wed — worked RAG demo end-to-end against bge-m3.

1 agent1 model
Details

Agents

  • doc_helper — model thinker · prompt doc_helper_sys · RAG aislang_docs

Models

  • thinker (chat) — prefer anthropic.claude_haiku_4_5 · fallback ollama.llama_3_1_8b

RAG knowledge

  • Vector store(s): aislang_docs
  • Corpus: aislang_overview.md, composition_primitives.md, k8s_deployment.md, rag.md
View source on GitHub

support-agent (paid)

single-file
MCP toolsDocker Composeevals

examples/support-agent-paid.ais — the v0.1 paid-tier demo variant.

1 agent1 model1 eval
Details

Agents

  • support — model fast · prompt system · tools docs

Models

  • fast (chat, tool_use) — prefer anthropic.claude_haiku_4_5 · fallback openai.gpt_4o_mini

RAG knowledge

  • Corpus: README.md, company-info.md
View source on GitHub

support-agent

single-file
MCP toolsDocker Composeevals

examples/support-agent.ais — the v0.1 canonical reference example.

1 agent1 model1 eval
Details

Agents

  • support — model fast · prompt system · tools docs

Models

  • fast (chat, tool_use) — prefer anthropic.claude_haiku_4_5, openai.gpt_4o_mini · fallback ollama.llama_3_1_8b

RAG knowledge

  • Corpus: README.md, company-info.md
View source on GitHub

support-agent-k8s

single-file
MCP toolsKubernetesevals

examples/support-agent-k8s.ais — Phase 2 Week 14 Tue.

1 agent1 model1 eval
Details

Agents

  • support — model fast · prompt system · tools docs

Models

  • fast (chat, tool_use) — prefer anthropic.claude_haiku_4_5, openai.gpt_4o_mini · fallback ollama.llama_3_1_8b

RAG knowledge

  • Corpus: README.md, company-info.md
View source on GitHub

skill-demo

single-file
skillsMCP tools

v0.3 Week 18 Wed-Thu — worked `skill { … }` block demo.

5 skills
Details

Skills

  • echo_greeting — echo 'hello from the aislang skill runtime' → string
  • describe_self — echo eyJydW50aW1lIjoiYWlzbGFuZyIsIm1vZGUiOiJzaGVsbCJ9 | base64 -d → json
  • list_cwd_files — ls -1 → list_string
  • evil_sudo — sudo cat /etc/shadow → string
  • read_notes — docs.tool.read_text_file → string

RAG knowledge

  • Corpus: notes.md
View source on GitHub

No agents match that capability.