This page consolidates already-stated beta boundaries from the current repo truth. Start with BETA_TESTING.md for the frozen public test tracks, and use mcp/integrations.md for the relay-vs-cloud planner/client contract.
| Beta Surface | Known Limitation | Workaround or Plan | Severity |
|---|---|---|---|
| Local daemon and local adapters | Canonical local proof is simulation-first; live adapter proof stays narrow, and the checked-in codex proof is still simulation-heavy rather than live-binary proven. |
Use the canonical local proof path from BETA_TESTING.md: ./scripts/smoke_test_v1.sh then make smoke. Treat /api/v1/compatibility and orchestratorctl instance --json as runtime truth for local readiness. |
friction |
| Local daemon MCP surface | Daemon-local /mcp/call is compatibility-only and is not the public remote planner MCP contract. |
Keep remote planners on relay /mcp or cloud /api/cloud/v1/mcp as described in mcp/integrations.md. |
blocker-for-use |
| Relay and daemon abort flows | Abort remains best-effort. Codencer only reports a successful abort when the active step actually reaches cancelled; it does not claim universal hard-kill semantics. |
Operate runs through the recorded state: inspect step state, result, validations, logs, gates, and artifacts after an abort request instead of assuming a forced stop. | friction |
| Relay artifact transport | Large binary artifact transfer is intentionally bounded; the connector is not a bulk file tunnel. | Use the normal artifact list and artifact content APIs for bounded evidence retrieval, and keep large-binary transfer expectations out of the relay path. | friction |
| Self-host planner authentication | Self-host planner auth is static bearer-token based and is not enterprise IAM. | Use the self-host relay/cloud paths as documented today, with the understanding that the current beta scope is narrow self-host operator use rather than enterprise identity integration. | friction |
| Generic MCP clients | Generic MCP clients beyond the manual JSON-RPC callers and official Go SDK helper remain expected-only; protocol behavior is proven, but specific client products are not universally repo-proven. | Prefer the repo-proven paths in mcp/integrations.md: relay/cloud HTTP, relay/cloud MCP, and the official Go SDK smoke helper. | friction |
| ChatGPT-style and Claude-style remote MCP paths | Product-specific ChatGPT-style and Claude-style integrations are compatibility-only in this beta contract. | Wire those clients only to relay /mcp or cloud /api/cloud/v1/mcp, keep the claim narrow, and use mcp/integrations.md as the Codencer-side contract. |
friction |
| Cloud runtime control | Cloud runtime HTTP and cloud MCP are only useful in composed mode, and cloud runtime connector ownership is explicit. A relay connector must still be claimed into org/workspace/project scope before cloud can use it. | Start codencer-cloudd with --relay-config or relay_config_path, then claim the runtime connector into tenant scope before using cloud runtime HTTP or MCP. |
blocker-for-use |
| Cloud connector enrollment | Connector enrollment-token issuance remains relay-config backed; cloud hosts connector ingress in composed mode, but it does not yet add a cloud-native enrollment-token lifecycle. | Issue enrollment tokens through the relay-backed self-host flow today. | friction |
| Docker self-host cloud baseline | make cloud-stack-smoke is the Docker baseline proof only. The Docker compose stack does not create a usable runtime instance by itself, and the broader runtime/MCP/SDK proof lives elsewhere. |
Use binary-native make cloud-smoke with composed-mode inputs when you need claimed runtime visibility, cloud runtime HTTP, cloud MCP, or official Go SDK proof. |
friction |
| Provider connectors overall | No provider is marketplace-approved, app-store-ready, OAuth-install ready, or vendor-depth complete in this phase. Live vendor-account proof is not part of the current beta promise. | Keep claims at the documented install/validate/action/webhook/audit depth from CLOUD_CONNECTORS.md, and use BETA_TESTING.md for the current provider track boundary. | friction |
| GitHub, GitLab, Jira, and Linear provider packaging | Slack is the strongest provider-shaped operator path today; GitHub, GitLab, Jira, and Linear remain narrower operator/package surfaces, and they do not yet have provider-specific binary smoke equivalent to Slack's current coverage. | Use Slack when you need the strongest current provider-shaped smoke path. For the other providers, rely on the generic cloud install/action routes and the provider-focused tests documented in CLOUD_CONNECTORS.md. | friction |
| Jira connector ingest | Jira is polling-first in the current beta track. Webhook ingest is explicitly deferred, and routed Jira webhook calls return 501 webhook_deferred instead of ingesting payloads. |
Use codencer-cloudworkerd for Jira polling and configure config.jql or config.project_key. Do not treat /webhook as the supported Jira ingest path in this phase. |
blocker-for-use |
| Cloud connector action history | There is still no public API route for listing action-log rows directly; action-log depth is mainly a store and audit truth improvement in this phase. | Use the existing cloud audit and event surfaces for operator visibility until a direct action-log listing route exists. | friction |
| WSL / Windows / agent-broker topology | The WSL/Windows layout is operator guidance, not an automated smoke-proof matrix, and the recommended topology keeps execution and artifacts on the Linux side. | Keep the repo checkout, daemon, connector, worktrees, and artifacts in WSL/Linux; keep agent-broker on Windows only when needed; inspect results through APIs and CLI rather than raw cross-side paths. |
friction |
| Ordered task execution | Ordered task execution remains wrapper-based and is not a native workflow engine inside Codencer core. | Use the checked-in wrapper examples and submit --wait --json loops for sequential execution. |
friction |