Basis

Brainbase is a platform for deploying and managing internal enterprise agents at scale. It provides a Postgres-like memory primitive that agents can use. Brainbase Labs describes an agent-improvement workflow based on defining an ontology, establishing benchmarks and evaluations, shipping a narrow first agent, and iterating through instructions, skills, and playbooks.

Agent Baseline is an open, versioned security resource for enterprise AI agents. It defines six security outcomes—Discover, Constrain, Authorize, Observe, Validate, and Respond—and 35 controls covering agent inventory and dependencies, scoped capabilities, identity and authority, correlated evidence, pre-use and ongoing validation, and revocation or quarantine. The project provides a white paper, source materials, implementation requirements, and evidence criteria for each control. It is maintained through open community review, issues, and implementation feedback, and serves as the baseline for the separate Secure Software Factory reference architecture.

Compound is an AI platform for buy-side finance teams that provides long-running, proactive deal-agent tasks. Its agents analyze data rooms, filings, and uploaded documents, then produce financial models, memos, and PowerPoint presentations with citations tracing figures to source files, pages, lines, model cells, or filing passages. Agents run tasks asynchronously and in parallel, including across thousands of files, with multiple agents and subagents able to work concurrently. Compound V2 provides multiplayer agents that teammates can share and direct through comments on workbooks and decks, along with browser-based Excel, PowerPoint, and Word editors built for agent-driven work, including formula-compatible spreadsheets and versioned changes. It supports PDFs, Excel files, CSV and TSV data, Word documents, PowerPoint files, images, and other common formats, and orchestrates models from multiple providers for different tasks. The service states that customer data is not used for training by default, is encrypted with AES-256 at rest and in transit, can be deployed in a VPC, and is SOC 2 Type II compliant.

agentacct is a local-first audit and observability tool for coding-agent work. It reads session logs from Claude Code, Codex, OpenCode, and Hermes, combines them with recorded tasks and test evidence, and produces a Work Receipt for each task covering commands run, files touched, tools used, time and token usage, cost, outcome, and evidence strength. Agent claims and machine verification are kept as separate axes, with verified status reserved for tasks whose live checks pass after the latest recorded work. The data is available through a macOS app, the live terminal dashboard launched with `agentacct tui`, and a local JSON API. State is stored in local files; the API listens only on loopback, and the project states that it has no login, cloud sync, or telemetry and does not request provider API keys.