# Archilas > Archilas gives context to the agents you already run. It keeps what was decided, who said it, when, and what an agent was told and did. Compatible with any MCP-capable AI. Archilas gives context to the agents you already run. It keeps what the team decided, who said it, when, what changed, and what an agent was told and did. It says it does not know when the context is not there. Memory is private to each user or team and is built only from their own material. Start free at https://www.archilas.com/signup?next=/app/connect, then open Connect. Book an early-access call at https://cal.com/archilas/archilas-intro. Free is 100 context deliveries a month, one integration, plus all AI chat context. Every paid plan includes all integrations. Slack, GitHub, Gmail, Notion, Linear, Google Calendar, Google Drive, and AI chats are connected. Compatible with any AI that supports MCP. Cursor, Claude Code, ChatGPT, and Codex are examples. Contact: hermes@archilas.com. X: https://x.com/archilas1 ## Home - [Archilas](https://www.archilas.com/): Context for the agents you already run ## Research - [Evidence in Context Is Not Enough: Why Small Language Models Fail at Long-Term Conversational Memory](https://www.archilas.com/research): Preprint by [OPERATOR NAME] on why small language models fail at long-term conversational memory even when the evidence is in the prompt. On one LoCoMo development split the preprint records 49 answer-stage failures where the gold evidence session was already in the prompt, and 22 retrieval-stage failures (14 ranked too low, 8 dropped by the context budget). It does not claim Archilas beats Mem0. - [PDF](https://www.archilas.com/research/evidence-in-context-is-not-enough.pdf): The same preprint as a PDF. ## Blog - [The small reader missed facts that were already in the prompt](https://www.archilas.com/blog/small-models-fail-at-agent-memory): I wrote a preprint on this. On one development split, most of the misses happened with the evidence already loaded. I am not claiming a win over Mem0. - [How I actually built the memory](https://www.archilas.com/blog/how-archilas-works): Sources come in. Notes keep the date and who said it. A newer decision marks the older one. An MCP client receives the current context. - [The agent was useful the second time, not the first](https://www.archilas.com/blog/how-memory-makes-agents-useful): The first session is a briefing. The second session is where memory either shows up or you paste the same thread again. - [The old decision was still in the notes](https://www.archilas.com/blog/stale-memory-current-fact): On 1 October we parked SSO. On 2 October we shipped it with the Team plan. Both notes exist. Only one is current. - [I did not want a store I have to query](https://www.archilas.com/blog/memory-agent-vs-memory-store): A store waits for a query. I wanted the agent to receive the current context when it starts work. - [Mem0, Zep, Letta, and what I am not claiming](https://www.archilas.com/blog/archilas-vs-mem0-zep-letta): They are memory systems you query from your own agent. I give the current context to the agent. There is no head-to-head score on this page. - [I use more than one AI in a day](https://www.archilas.com/blog/persistent-memory-for-everyday-ai): The memory should not live inside one chat product. I connect the tools once and point each MCP client at the same server. - [I would rather the agent say it does not know](https://www.archilas.com/blog/memory-that-does-not-fabricate): A fluent sentence with no note behind it is the bug I am trying not to ship. - [MCP is the socket, not the memory](https://www.archilas.com/blog/what-is-mcp): MCP is how a client calls a tool. Archilas is the tool that hands back the current context. - [The context died when I closed the tab](https://www.archilas.com/blog/why-agents-lose-context): The model did not forget. The session ended. The next tool opened with an empty prompt. - [Search found the paragraph. It did not know which decision won](https://www.archilas.com/blog/ai-memory-vs-rag): RAG is how I pull a passage. It is not how I know the October decision replaced the September one. ## Resources - [Resources](https://www.archilas.com/resources): Writing on memory, agents, and the preprint. - [Blog](https://www.archilas.com/blog): All posts. ## Contact - [Email [OPERATOR NAME]](mailto:hermes@archilas.com): hermes@archilas.com - [X](https://x.com/archilas1): @archilas1 - [Start free](https://www.archilas.com/signup?next=/app/connect): Create a workspace, then open Connect. Free is 100 questions a month, one integration, plus all AI chat context. - [Book an early-access call](https://cal.com/archilas/archilas-intro): Intro call. - [Pricing](https://www.archilas.com/#pricing): Free (100 questions, 1 integration), Pro ($19/month, 1,000 questions, 1 seat), Builder ($49/month, 2,500 questions, 1 seat), Studio ($99/month, 5,000 questions, 3 seats), and Team ($199/month, 10,000 questions, 10 seats). Every paid plan includes all integrations. No Enterprise tier. Checkout in the app is off. - [Docs](https://www.archilas.com/docs): MCP for any compatible AI, connectors, memory, and the HTTP API. - [Compare](https://www.archilas.com/compare): Search-and-paste versus Archilas. No general benchmark claim. - [Security](https://www.archilas.com/security): Workspace separation, AES-256-GCM connector credentials, no certification claimed. ## Optional - [Privacy](https://www.archilas.com/privacy): Controller [OPERATOR NAME], data categories, subprocessors, and rights. - [Terms](https://www.archilas.com/terms): Terms. Lemon Squeezy is the merchant of record for purchases. - [Cookies](https://www.archilas.com/cookies): Essential cookies, and analytics only after consent.