Open-source tools that AI agents use as infrastructure. Pure TypeScript, zero native dependencies, MCP-first, event-driven, crash-safe by default.

AI-first embedded database for LLM agents. Zero native dependencies, pure TypeScript.

• MCP tools — CRUD, schema lifecycle, search, archive, vector search, blob storage, real-time subscriptions
• Persisted schemas — defineSchema() with typed fields, validation, defaults, hooks, auto-indexing; agent-discoverable at runtime via db_get_schema
• Vector search — HNSW nearest neighbor with 6 embedding providers
• Blob storage — attach files to records, works on filesystem and S3
• Crash-safe — append-only WAL with immutable snapshots

npm install @backloghq/agentdb

Persistent task management for Claude Code. Tasks survive sessions so work started by one agent can be picked up by another.

• MCP tools — full task lifecycle: add, modify, done, delete, start, stop, annotate, archive
• Skills — /backlog:plan, /backlog:tasks, /backlog:standup, /backlog:refine, /backlog:spec, /backlog:implement, /backlog:handoff
• Agent coordination — tasks auto-assign to spawned agents, sync with Claude's built-in task system

/plugin marketplace add backloghq/backlog

Embedded event-sourced document store. The engine under AgentDB.

• Append-only WAL — every mutation recorded as an operation, state derived by replay
• Immutable snapshots — fast cold-start without replaying the full log
• Multi-writer — Lamport clocks for last-writer-wins across concurrent agents
• Tradeoff — a crash can lose buffered ops (up to 100ms). Default "immediate" mode is safe.

npm install @backloghq/opslog

S3 storage backend for opslog. Enables multi-writer concurrency across machines via Amazon S3.

npm install @backloghq/opslog-s3

Log-structured full-text search index. Segment-based posting lists with LSM compaction, BM25 ranking, zero native dependencies.

• Size-tiered LSM compaction — fanout-based merge; write amplification bounded at O(N log_fanout N)
• BM25 ranking — configurable k1/b, top-k heap
• Streaming segment writes — atomic commits via manifest swap; crash-safe orphan recovery
• Pluggable storage — local FS by default, S3 via @backloghq/termlog-s3
• Scales — 1M+ docs per index without per-file size cliffs

npm install @backloghq/termlog

S3 storage backend for termlog. Streaming segment writes via S3 multipart upload; supports Amazon S3, MinIO, and LocalStack.

npm install @backloghq/termlog-s3

Documentation · GitHub