Cortex
An open-source API knowledge-layer tool that treats API specifications (OpenAPI, AsyncAPI, GraphQL, gRPC, OpenRPC) as the single source of truth. It automatically derives and generates interactive documentation, multi-language typed SDKs, and MCP servers, unifying repetitive development work across the API lifecycle.
Timeline (last 90 days)
Cortex proposes "memory consolidation": distilling the gains brought by context into low-rank or FFN parameter updates merged into the model itself, recovering 88.9% of the context gains after 50 sequential fact edits while retaining 99.8% of the original performance.
Mitosis Labs' research page reports that Cortex (its memory system) achieved 96% accuracy, 0.0% error rate, and an average response time of 4.4 seconds in the Agentic Memory Index's 272-task evaluation.
Cortex's benchmark results describe a memory-system evaluation, are not measured results from a Private SuperIntelligence customer deployment, and are not a comparison with ChatGPT Enterprise.
Cortex is an open-source project that uses the API specification as the single source of truth and automatically derives documentation, SDKs, and MCP servers from it.
Cortex was built by Nick Chisiu.
Cortex supports five types of API specifications as input formats: OpenAPI, AsyncAPI, GraphQL, gRPC, and OpenRPC.
Cortex derives three types of artifacts from API specifications: interactive documentation, typed SDKs, and MCP servers.
The typed SDKs generated by Cortex cover 11 programming languages.
By including MCP servers as a default output, Cortex differentiates itself from traditional API code generators such as Swagger Codegen and OpenAPI Generator, serving both human developers and AI consumers.
Typed SDKs use the language's native type system to solidify API request and response structures, and regenerating after a specification file update synchronizes all language versions.
13 more timeline events
All Facts (20)
Cortex proposes "memory consolidation": distilling the gains brought by context into low-rank or FFN parameter updates merged into the model itself, recovering 88.9% of the context gains after 50 sequential fact edits while retaining 99.8% of the original performance.
50%UnverifiedMitosis Labs' research page reports that Cortex (its memory system) achieved 96% accuracy, 0.0% error rate, and an average response time of 4.4 seconds in the Agentic Memory Index's 272-task evaluation.
50%UnverifiedCortex's benchmark results describe a memory-system evaluation, are not measured results from a Private SuperIntelligence customer deployment, and are not a comparison with ChatGPT Enterprise.
50%UnverifiedCortex was built by Nick Chisiu.
50%UnverifiedCortex is an open-source project that uses the API specification as the single source of truth and automatically derives documentation, SDKs, and MCP servers from it.
50%UnverifiedCortex supports five types of API specifications as input formats: OpenAPI, AsyncAPI, GraphQL, gRPC, and OpenRPC.
50%UnverifiedCortex derives three types of artifacts from API specifications: interactive documentation, typed SDKs, and MCP servers.
50%UnverifiedThe typed SDKs generated by Cortex cover 11 programming languages.
50%UnverifiedBy including MCP servers as a default output, Cortex differentiates itself from traditional API code generators such as Swagger Codegen and OpenAPI Generator, serving both human developers and AI consumers.
50%UnverifiedTyped SDKs use the language's native type system to solidify API request and response structures, and regenerating after a specification file update synchronizes all language versions.
50%UnverifiedAs an open-source project, Cortex allows teams to self-host and review the generation logic, avoiding hosting API specifications with third-party closed-source services.
50%UnverifiedThe actual generation quality of Cortex still needs to be verified through real-world testing, including whether the SDKs conform to idiomatic usage in each language, MCP server reliability, and the degree of support for complex API specifications.
50%UnverifiedCortex converts information into typed claims with explicit semantic types rather than unstructured text fragments.
50%UnverifiedWhen contradictory information arises, Cortex tracks the contradiction rather than overwriting the old memory.
50%UnverifiedCortex claims to complete integration with Claude in 2 minutes and is compatible with any MCP client.
50%UnverifiedCortex is deployed on EU infrastructure and complies with data compliance requirements such as GDPR.
50%UnverifiedCortex's core engine is patent-pending.
50%UnverifiedCortex is a hosted semantic memory service, allowing developers to avoid building their own vector databases or maintaining memory pipelines.
50%UnverifiedAccording to official data, approximately 80% of write requests in production environments are judged by Cortex to be redundant and directly rejected.
50%UnverifiedCortex was independently developed by a single independent developer, Filippo Pilotta, from Bergamo, Italy.
50%