Bilinc

Verifiable state plane for autonomous agents.

pip install bilinc

Most agent memory systems are a vector store with a wrapper. Bilinc is a state plane: every belief can be verified before it lands, logically revised when it conflicts, and audited so an agent can explain what changed and why.

Others store memories. Bilinc manages truth.

Why Bilinc

Long-running agents fail in predictable ways:

• They store contradictions they never catch.
• They overwrite useful context with noisy recency.
• They cannot roll back when memory or tool state goes bad.

Bilinc gives agents a state layer with verification, belief revision, provenance, and rollback built in.

Core capabilities

Quick start

from bilinc import StatePlane
from bilinc.storage.sqlite import SQLiteBackend

backend = SQLiteBackend("agent_state.db")
sp = StatePlane(backend=backend, enable_verification=True, enable_audit=True)

# Commit a belief: verification and audit happen before write.
sp.commit_sync(
    "user_preference",
    {"theme": "dark", "language": "python"},
    memory_type="semantic",
    importance=0.9,
)

# Recall across FTS5, vector, and graph signals.
results = sp.recall_all_sync("user preference", limit=5)

# Snapshot, mutate, rollback.
snapshot_id = sp.snapshot_sync()
sp.commit_sync("user_preference", {"theme": "light"})
sp.rollback_sync(snapshot_id)

CLI

bilinc --db ./agent.db commit --key USER_PREF --value '{"theme": "dark"}'
bilinc --db ./agent.db recall --key USER_PREF
bilinc --db ./agent.db forget --key USER_PREF
bilinc --db ./agent.db status

# Hermes integration
bilinc hermes bootstrap
bilinc hermes smoke

MCP integration

Bilinc ships as an MCP server for Claude Code, Cursor, and any MCP-compatible agent:

{
  "mcpServers": {
    "bilinc": {
      "command": "python",
      "args": ["-m", "bilinc.mcp_server.server_v2"],
      "env": { "BILINC_DB_PATH": "~/bilinc.db" }
    }
  }
}

MCP tools include: commit_mem, recall, revise, forget, consolidate, contradictions, diff, snapshot, rollback, status, verify, query_graph, bilinc_recall_smart, bilinc_query_analysis, bilinc_event_segment, bilinc_summarize, bilinc_health, bilinc_benchmark, bilinc_export, and bilinc_import.

LangGraph checkpointing

Use Bilinc as a verified persistent checkpoint store for LangGraph agents:

from langgraph.graph import StateGraph
from bilinc import StatePlane
from bilinc.storage.sqlite import SQLiteBackend
from bilinc.integrations.langgraph import LangGraphCheckpointer

sp = StatePlane(backend=SQLiteBackend("checkpoints.db"), enable_verification=True)
checkpointer = LangGraphCheckpointer(sp)

graph = StateGraph(...).compile(checkpointer=checkpointer)

Every checkpoint can flow through Bilinc's revision and verification pipeline, making long-running LangGraph state inspectable and rollback-capable.

Architecture

StatePlane
├── WorkingMemory          PFC-inspired active slots and eviction
├── AGM Engine             EXPAND / CONTRACT / REVISE
├── Dual-Process Arbiter   fast path + deliberate verification path
├── StateVerifier          Z3 SMT contradiction gate
├── AuditTrail             Merkle chain and provenance
├── KnowledgeGraph         entities, relations, spreading activation
├── Hybrid Recall          FTS5 → vector → KG → RRF fusion
├── ContextBudgetRL        adaptive token allocation by memory type
├── Storage                SQLite / PostgreSQL
└── MCP Server v2          stdio + authenticated HTTP

Benchmarks

These are repository benchmark runs, not hosted-service claims. Reproduce them from the benchmark scripts and fixtures in this repo.

Comparison

Installation

# Core
pip install bilinc

# PostgreSQL backend
pip install "bilinc[postgres]"

# HTTP MCP server
pip install "bilinc[server]"

# Development
pip install -e ".[dev]"
pytest tests/ -v

License

BSL 1.1 — free for personal and research use. Commercial SaaS use is restricted until 2030, then Apache 2.0.