Crate agentfs 

AgentFS - Filesystem Abstraction for AI Agents

AgentFS provides a high-level filesystem abstraction for AI agents, offering POSIX-like file operations, key-value storage, and tool call auditing.

Features

• Filesystem: POSIX-like file and directory operations
• KV Store: Key-value storage for agent state
• Tool Recording: Audit trail for agent tool calls
• Backend Agnostic: Works with any AgentDB backend (SQL, KV, Graph)

Example

use agentfs::AgentFS;
use agentsql::{SqlBackend, SqlBackendConfig};

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    // Create a SQLite-backed agent filesystem
    let backend = SqlBackend::sqlite("agent.db").await?;
    let agent_fs = AgentFS::new(Box::new(backend), "my-agent", "/agent").await?;

    // Write a file
    agent_fs.fs.write_file("/output/report.txt", b"Hello World").await?;

    // Store state
    agent_fs.kv.set("session_id", b"12345").await?;

    // Record a tool call (workflow-based API)
    let id = agent_fs.tools.start(
        "web_search",
        Some(serde_json::json!({"query": "Rust async"}))
    ).await?;
    // ... perform the search ...
    agent_fs.tools.success(id, Some(serde_json::json!({"results": []}))).await?;

    Ok(())
}

Re-exports

pub use error::AgentFsError;

pub use error::Result;

pub use filesystem::DbFileSystem;

pub use filesystem::FileSystem;

pub use filesystem::Stats;

pub use kvstore::DbKvStore;

pub use kvstore::KvStore;

pub use tools::DbToolRecorder;

pub use tools::ToolCall;

pub use tools::ToolCallStats;

pub use tools::ToolCallStatus;

pub use tools::ToolRecorder;

Modules

error

Error types for AgentFS

filesystem

Filesystem operations for AgentFS

kvstore

Key-value store for agent state

rig_integration

Rig.rs integration documentation (available even when feature is disabled) Rig.rs FileSystem trait integration for AgentFS

tools

Tool call recording and auditing

Structs

AgentFS

Main AgentFS struct providing filesystem, KV store, and tool recording