AI agent infrastructure 8

Behavioral APIs for humanlike AI agents Humalike is a behavioral infrastructure platform for humanlike AI agents. It provides APIs and research-backed components that help agents know when to speak, read social cues, remember people, adapt to group norms, and behave more naturally in 1:1 and group conversations.

PostgreSQL platform for hybrid AI agent memory and retrieval Polygres is an all-in-one PostgreSQL platform for AI agents that combines relational data, graph traversal, vector search, and hybrid retrieval through a unified API. It turns existing PostgreSQL databases into working memory for agents, allowing them to retrieve structured rows, connected relationships, semantic matches, full-text results, and filtered context without using a separate vector database, graph database, or synchronization pipeline. Polygres includes pgGraph for foreign-key-based graph traversal and pgContext for fused retrieval across dense HNSW vectors, sparse search, full-text search, filtering, recommendations, discovery, grouped results, and lookups. Users can deploy it as a managed cloud database or self-host its open-source components.

Versioned filesystem for persistent, collaborative AI agent workspaces Mesa is a versioned filesystem and virtual filesystem platform built for AI agents. It allows agents to access persistent files, create isolated branches, run parallel agent swarms, track changes, request human approvals, and roll back work through one API. Mesa supports POSIX-compatible filesystem access, large files, shared memories, documents, code, datasets, models, and media. It provides sub-50ms reads and writes through MesaFS, instant workspace forks, strong consistency, fine-grained access controls, and enterprise deployment options.

Governed infrastructure for building AI agents on connected business data CoreBase is a governed AI agent infrastructure platform for building and deploying customer-facing or internal AI agents on live business data. It connects databases, REST and GraphQL APIs, SaaS applications, on-premise systems, and legacy databases through an OpenAI-compatible API, SDK, or embeddable chat widget. CoreBase includes permissions, source scoping, read-only defaults, approval gates, audit logs, tenant isolation, prompt-injection screening, cost controls, and support for scheduled or event-driven automations. Its open-source CoreMCP bridge enables secure outbound-only access to on-premise, air-gapped, and legacy systems.

Composable infrastructure for building, running, and deploying AI agents. Rivet is infrastructure for building, operating, automating, and deploying production-grade AI agents. Its composable stack includes Actors for durable processes, real-time communication, streaming, and memory; agentOS for secure code execution, filesystems, isolation, and session transcripts; Workflows for restart-safe multi-step operations; and Dynamic Apps for deploying applications created by agents. Rivet can run on Rivet Cloud, in a private VPC, on-premises, in air-gapped environments, or on custom infrastructure.

Cited geospatial data API for AI agents and physical-world decisions Mireye is a geospatial infrastructure platform that connects AI agents to authoritative physical-world and environmental data through one API and MCP server. It enables agents to ask location-based questions, geocode addresses, retrieve cited geographic fields, measure proximity and driving distances, and request new data fields. Responses include structured values, source attribution, source URLs, fetch timestamps, confidence scores, and typed refusals when data quality is insufficient. Mireye covers US geospatial data from sources such as USGS, FEMA, NOAA, USDA, EPA, EIA, FCC, Census, NREL, and other federal, county, open, and licensed datasets. It supports applications including data center siting, renewable energy development, insurance underwriting, commercial lending, mortgage diligence, land evaluation, and infrastructure planning.

Cloud Linux VMs for developers, AI agents, and production workloads exe.dev is a cloud platform for developers and AI agents that provides instant, persistent, disposable, and shareable Linux virtual machines. Users can launch VMs through SSH, access them through HTTPS, install software with apt, manage services with systemd, and run coding agents such as Shelley. The platform supports development environments, AI agent sandboxes, data analysis, VPS hosting, team collaboration, and production infrastructure. VMs include configurable CPU, memory, and disk resources, with private-by-default access and IAM-based sharing.

AI token optimization stack for reducing agent costs and context Caveman is an AI efficiency operating stack that reduces the cost and context size of agent-native development. It provides an MIT-licensed skill, a free local proxy, an Agent SDK, and planned cloud and enterprise services. Caveman compresses logs, JSON, code, diffs, tables, tool outputs, and files before provider calls while preserving recoverable original bytes. It also supports caching, eval-gated model routing, token usage visibility, savings estimates, and verification workflows. The local wrap works with users' own provider keys and does not require a Caveman account.