1 / 3
【Overview】 Honcho is open-source AI memory for agents. It stores messages and runs reasoning to build representations of users and other peers, then returns token-budgeted context so agents keep continuity across sessions, tools and machines. Instead of starting every session cold, an agent can retrieve what it already knows about a person and carry that understanding forward. The problem it addresses is statelessness. Agents that forget the previous conversation are hard to build into companions, tutors, support bots or game characters. Honcho keeps a persistent, reasoned representation of each peer and turns it into context that fits inside a token budget. It is built for developers creating stateful AI agents, users of coding agents such as Claude Code, Codex, OpenCode and Kilo Code, and teams working on AI companions, gaming NPCs, education tutors, customer support and productivity tools. Startups with less than $5M raised building memory-dependent apps are explicitly targeted. 【Key features】 • context(): Retrieves formatted conversation context from sessions, combining a summary and recent messages within a token budget, convertible to OpenAI and Anthropic formats. • Peer chat / dialectic reasoning: Answers queries about a peer or across all peers by reasoning over stored messages and conclusions, in tiers from Minimal to Max. • Semantic search over conclusions: Fetches semantically relevant conclusions for a query string via search_query with distance and top-k controls. • Dreaming: Runs asynchronous background reasoning to optimize understanding of each peer without impacting runtime performance. • Scoped perspectives / Peers: Models users, agents, NPCs and groups as peers that can be added to sessions and scoped to global or narrow perspectives. • Session-scoped representations: Limits retrieved conclusions to the current session via limit_to_session. • MCP server: Gives any MCP-compatible client persistent memory via https://mcp.honcho.dev. 【Use cases】 1. Give an AI agent persistent memory of its users across sessions 2. Track misconceptions and learning progress in tutoring agents 3. Persist customer history across support sessions, channels and agents 4. Give NPCs opinions, relationships and identities in games 5. Share memory across a chat companion, coding agent and ingestion jobs 【Pricing】 Honcho Memory - Ingestion: $2.00/M — Store + Neuromancer reasoning Startups (<$5M raised): $1,000 — $1,000 in credits, 12 months subsidized pricing, Integration support Enterprise: Custom plans — Forward-deployed engineers, Dedicated integration and maintenance support 【Pros and cons】 Pros: - Developers who need agents to remember users across tools and machines get a purpose-built memory layer rather than wiring up their own store and summarization. - Teams that already run MCP-compatible clients can add persistent memory through the hosted MCP server without building a custom integration. Cons: - The context() call is capped by its tokens parameter, so retrieved context must fit a budget. - The Max tier of peer chat / dialectic reasoning costs $0.50/q and runs asynchronously. - Scope settings require peer_target plus a workspace-level or admin-level key, and are mutually exclusive with peer_perspective. 【FAQ】 Q: How is Honcho priced? A: Ingestion is billed at $2.00/M for storing messages plus Neuromancer reasoning. Startups with less than $5M raised can get $1,000 in credits, 12 months of subsidized pricing and integration support, and Enterprise custom plans are available. Q: Who is Honcho built for? A: It targets developers building stateful AI agents, users of coding agents like Claude Code, Codex, OpenCode and Kilo Code, and teams building AI companions, gaming NPCs, education tutors, customer support and productivity tools. Q: Can Honcho keep memory across different tools and machines? A: Yes. It stores messages and builds peer representations so an agent keeps continuity across sessions, tools and machines, and any MCP-compatible client can get persistent memory through https://mcp.honcho.dev.
Honcho Memory - Ingestion · $2.00/M (Store + Neuromancer reasoning) Startups (<$5M raised) · $1,000 ($1,000 in credits, 12 months subsidized pricing, Integration support) Enterprise (Custom plans, Forward-deployed engineers, Dedicated integration and maintenance support)
WPS AI is an AI assistant built into WPS Office that helps users write documents, generate presentation outlines and…
Baidu Comate is an AI coding assistant built by Baidu on its ERNIE model. It goes beyond simple autocomplete: it…
UniConverter is an AI media conversion and enhancement tool for video, audio, and images. It combines everyday…
Added
10/20/2026
Views
843
Clicks
4
6 similar tools to consider instead of Honcho