Sutra MCP Gives AI Agents Portable, User-Controlled Memory Across Sessions

·1 views

A developer has built Sutra MCP, a hosted service that provides AI agents with persistent, portable memory and reusable skill libraries via the Model Context Protocol. The tool addresses a common frustration where context and decisions are lost when switching between AI agents or chat sessions. Users can deliberately save observations, search relevant memories, and organise work into named project clusters, keeping personal memory separate from reusable task workflows. Sutra supports both OAuth 2.1 and API key authentication and is compatible with clients such as Cursor and OpenCode through a single Streamable HTTP endpoint. The service is listed in the official MCP Registry and allows users to export their full memory dataset for portability and control.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Google Limits AI Security Model to 650 Orgs; Developers Build DIY Alternatives

Google announced Gemini 3.8 Flash Cyber on Tuesday, a security-focused AI model that helped its Cloud Vulnerability Research team find a critical vulnerability in under two hours — work that typically takes months. The model is restricted to roughly 650 organizations globally through a program called Fairwind, covering governments, critical infrastructure operators, and select partners such as CrowdStrike and Palo Alto Networks. Independent developers and startups have no defined path to access the tool, which shares its architecture with the publicly available Gemini 3.8 Flash model but includes more permissive settings for cybersecurity tasks. On the same day, Anthropic and OpenAI also launched or expanded their own restricted security AI programs, signaling a coordinated industry move toward tiered access for powerful security-capable models. In response to the access gap, some developers are building alternative vulnerability-scanning pipelines using the public Flash model, which is priced at $0.75 per million input tokens through the end of 2026.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Why the Cheapest Game Dev Outsourcing Partner Often Costs You More

Studios evaluating game development outsourcing typically compare hourly or daily rates, but this reflects only a fraction of the true cost involved. Hidden expenses such as communication overhead, ramp-up time, timezone friction, quality variance, and internal management burden can significantly erode savings from a lower-rate provider. UK-based co-development studio Ocean View Games draws on experience from both sides of outsourcing relationships to highlight these overlooked cost drivers. The firm argues that external teams often require more detailed documentation and explicit acceptance criteria due to limited product context, adding further internal workload. A more meaningful comparison, it contends, is the total cost of getting work integrated, accepted, and maintainable — not just the quoted rate.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

OB1 AI Stack Impresses but Demands Careful Docker Network Configuration

A developer reviewed the open-source OB1 project, which consolidates AI memory, a gateway, and a chat interface into a single self-hosted stack instead of multiple SaaS tools. The architecture was found to be straightforward and appealing for teams that want to keep prompts and API routing within a private network. A key issue emerged during Docker deployment: using localhost as the gateway URL inside a container causes a connection-refused error, since localhost resolves to the container itself rather than the host. The fix involves referencing the Compose service name directly and ensuring both containers share the same internal network. Beyond networking, the reviewer cautioned that database credentials, port bindings, and log retention policies remain the adopting team's responsibility before any public exposure.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How to Implement Advanced Server-Side Caching in Next.js Using Redis and ISR

A technical guide published on DEV Community outlines advanced server-side caching strategies for Next.js applications to address performance bottlenecks caused by redundant computations and repeated database queries. The article describes a layered caching architecture spanning the client browser, CDN edge, Next.js server cache, Redis, and the database or API layer. It explains how tools like Redis, Incremental Static Regeneration (ISR), and HTTP cache headers work together to reduce latency and improve scalability beyond basic Next.js data-fetching methods. Practical code examples demonstrate how to set up a Redis client and cache API responses for a configurable duration, such as five minutes. The guide targets developers building large-scale Next.js applications who need caching solutions suited to dynamic or frequently changing content.

0 comments Read more at DEV Community