v0.5.0 — 500+ Tool Packs

Build agents with structural trust

Design the constraints of intelligence, not the intelligence itself. A Go runtime where trust is the product.

Get Started → Documentation
go get github.com/felixgeelhaar/agent-go
intakeexploredecideactside effectsvalidatedonefailed
current: intake
step 1/8
// CAPABILITIES

Engineered for Production

◈

Canonical State Machine

7 states with explicit semantics. Side effects only in Act state. Deterministic, auditable execution flow.

⬡

Policy Enforcement

Tool eligibility per state. Budget limits. Approval requirements for destructive actions. Constraints, not hope.

◇

Resilient Execution

Circuit breakers, retry with backoff, bulkheads, timeouts. Built on battle-tested patterns via Fortify.

▣

Tool Annotations

ReadOnly, Destructive, Idempotent, Cacheable. Metadata that influences behavior, not just documentation.

◉

Human-in-the-Loop

AskHuman decision type. Pause execution, collect input, resume with evidence. Built-in approval workflows.

⬢

Knowledge Store

Vector embeddings with cosine similarity search. RAG-ready architecture for persistent agent memory.

// TOOL PACKS

500+ Tools, Opt-In

Modular contrib packages. Import only what you need. Every tool has annotations, schema validation, and resilience built in.

🔐

Security & Crypto

Encryption, hashing, signing, key generation, JWT operations

19 tools
📊

Data Processing

JSON, YAML, CSV, Excel, archive, template, regex

80+ tools
🌐

Network & Protocol

HTTP, DNS, SSH, MQTT, gRPC, WebSocket, port scanning

90+ tools
🗄️

Database & Storage

SQL, PostgreSQL, SQLite, Redis, filesystem operations

30+ tools
🧠

AI/ML & Analytics

Embeddings, classification, NER, statistics, charting

90+ tools
🔗

Integrations

GitHub, Slack, Jira, browser automation, Git operations

50+ tools
📝

Text & Content

Markdown, HTML, XML, QR codes, text similarity, AST parsing

80+ tools
🛠️

Utility

UUID, hash, color, random, validation, date/time, cron, semver

90+ tools
go get github.com/felixgeelhaar/agent-go/contrib/pack-json@v0.5.0
// STATE MACHINE

Canonical Execution States

State Purpose Side Effects Terminal
intake Normalize and validate goal No No
explore Gather evidence, read-only operations No No
decide Plan next action No No
act Execute side effects Yes No
validate Confirm action outcomes No No
done Successful completion No Yes
failed Error termination No Yes
◈

Key Invariant: Only the act state permits side effects. All other states are pure observation and decision.

// GET STARTED

Quickstart

terminal
# Add agent-go to your project
go get github.com/felixgeelhaar/agent-go
// SEE IT IN ACTION

Agents in Action

Watch real agents execute through the state machine. Click the tabs to explore different use cases.

File Operations Agent

View Code

Manages files in a workspace: read, write, list, and delete operations.

Available Tools

list_dirReadOnly

List files and directories

read_fileReadOnly

Read contents of a file

write_fileIdempotent

Write content to a file

delete_fileDestructive

Delete a file (requires approval)

Execution Trace

intake
explore
decide
act
validate
done
intakeGoal received
Create a hello.txt file with a greeting message
⏸ Click to pause
step 1/6
// ARCHITECTURE

Domain-Driven Design

domain/

Core Domain

  • agent/ State, Decision, Run, Evidence
  • tool/ Tool, Annotation, Registry
  • policy/ Budget, Approval, Constraint
  • ledger/ Audit trail, Events
application/

Application Layer

  • engine.go Orchestration service
  • options.go Configuration
infrastructure/

Infrastructure

  • statemachine/ Statekit integration
  • resilience/ Fortify patterns
  • planner/ Mock, Scripted, LLM
  • storage/ Memory, Filesystem
interfaces/api/

Public API

  • agent.go Engine, Options, Re-exports
  • builders.go Fluent constructors
contrib/

Tool Packs new

  • pack-*/ 500+ opt-in tools as Go modules
  • storage-*/ Postgres, Redis, SQLite, DynamoDB
  • planner-llm/ Claude, OpenAI, Gemini, Ollama
  • dashboard/ Observability UI

Ready to build trustworthy agents?

Start with constraints. Add intelligence later.

Read the Docs → View on GitHub