Agent Atlas / Runtimes and workbenches

openclaw/openclaw

An agent runtime whose sessions, memory, command approval, scheduled wake-ups, audit trails, and channel-facing controls make long-lived operation the central teaching surface.

Agent runtime Scope profile Reported + sourced Reviewed August 26, 2026
Intended operator

People running persistent agents across terminal, messaging, and scheduled work.

Primary surfaces

agent runtime / sessions and channels / scheduled operations

Tracked repository

openclaw/openclaw

Teaching mechanism

A generated diff needs a reproducible base

Diff evidence becomes trustworthy only when the runtime records enough repository state to distinguish agent output from a guessed comparison.

Evidence limitA better diff base does not prove the patch is correct or that every downstream review preserves the provenance.
Inspect the source
System map status

Reporting is still in progress.

This scope profile names the project's place in the field and one useful mechanism. It does not claim that every runtime boundary has been inspected.

Learn Through This Project

10 lessons
Chapter 1 Before the First Model Request Chapter 2 The Invisible Prompt Chapter 3 Tool Calls: Contract, Authorization, Execution Chapter 4 Loops, Workflows, and Recovery Chapter 5 Delegation Is a Distributed System Chapter 6 Memory Is a Typed Storage Contract Chapter 7 The Approval Trap Chapter 8 The Human Control Loop Chapter 9 Trust Must Make Its Case Chapter 10 The Future Has Three Branches

Compare Its Boundaries

8 labs
Comparison labTool ContractsComparison labGoverned MemoryComparison labDelegation ReceiptsComparison labApproval BoundariesComparison labHuman WorkbenchComparison labSessions And RecoveryComparison labEvidence And EvaluationComparison labMaturity Signals

Latest Reporting

8 articles
Who Supplies the Agent? OpenAI Hosts It, Pydantic Assembles It

September 28, 2026 — Pydantic’s weekend changes tie a framework, a coding harness and a terminal client to one release process. Set beside OpenAI’s managed Agents API, they reveal a choice that model rankings miss: which parts of the agent your team still has to run. openai/codex, pydantic/pydantic-ai, openclaw/openclaw

 OpenClaw Retires the Ledger That Watched Its Other Ledgers

September 28, 2026 — Tasks and TaskFlow are gone from OpenClaw’s main branch. Background execution remains, but plugins lose a shared interface—and some upgrade scenarios still lack a completed proof. openclaw/openclaw

 OpenClaw Had Stopped the Child. Restart Brought Back the Parent.

September 20, 2026 — A waiting conversation could survive its own cancellation in OpenClaw’s recovery records. A new repair saves the parent’s stopped state before answering the user—and protects a newer turn from that delayed write. openclaw/openclaw, NousResearch/hermes-agent

 A Shared Terminal Is Not Shared Permission

August 26, 2026 — OpenClaw’s new terminal-input path separates an agent being allowed to see an operator’s terminal from being allowed to send a keystroke into it. openclaw/openclaw

 A Release Should Not Fetch Its Rules From a Moving Main

August 22, 2026 — OpenClaw is making a narrow release promise: when the product is frozen, the workflow allowed to publish it must stay identifiable even after main changes. openclaw/openclaw

 When a Context Limit Drops the User’s Turn

August 17, 2026 — OpenClaw found a required Codex compaction preflight treating an intentional no-op as a fatal failure—so an overlong session could lose the incoming user turn before the model ever answered. openai/codex, openclaw/openclaw

 OpenClaw’s shared channels gain a per-requester login

August 12, 2026 — A new opt-in OAuth mode for HTTP MCP servers lets a trusted sender connect their own account instead of letting the whole room run on one operator credential. openclaw/openclaw

 QM is trying to turn the personal agent into company infrastructure

August 1, 2026 — Y Combinator's new open-source harness does not win its case by running Codex or Claude in Slack. Its more consequential bet is that every person and room needs a clearly owned memory, workspace, permission set, schedule, and durable computer. yc-software/qm, block/buzz, egregore-labs/egregore, openclaw/openclaw, NousResearch/hermes-agent, crewAIInc/crewAI, microsoft/agent-framework

Primary Sources

OpenClaw repository

Primary public runtime source.

Open Questions

  • Which background effects remain attributable after a session sleeps, wakes, migrates, or exits?

Evidence state describes what this Atlas profile can support today. It is not a product score and it is not evidence that uninspected capabilities are absent.