How can we help?
Guides for every screen, fixes for the errors people meet first, and the reference behind it all.
- UnderstandLearn the conceptsNew to objectives, requirements, constraints or stories? What each one is, what goes wrong without it, and why AI agents need it.Start learning
- Get things doneGuides by role and topicStep-by-step guides for every screen and task, from connecting your agent to planning work on boards.Find a guide
- Fix a problemWhen something goes wrongSituations people meet with their agents, error messages and quick fixes.Find a fix
About this help center
This help center explains how to build with coding agents when Corpole keeps the record of what binds. It is written for everyone who touches a project: the founder who builds a product idea with an agent, the person who builds alone with Claude Code, the developer in a team, the product owner who decides what the product must do, the lead who plans and keeps delivery disciplined, the tester, the admin, and the reviewers who evaluate or secure it.
It is organised three ways, so you can come in from wherever you are. By role: each role has a Start here page with five steps and the articles that matter most to it. By situation: “When things go wrong” describes the problems people actually meet with agents — a forgotten decision, a rewritten requirement, a looping test, a production outage — and what to do about each. By topic: guides for every screen, how-to articles for common tasks, concepts that explain why Corpole works the way it does, fixes for error messages, and complete reference lists. If the words themselves are new — objective, business requirement, epic, acceptance criterion — the Learn section explains each one from the ground up.
Every article describes what Corpole does today; nothing here is a promise about a future release. The screen guides are the same text you see in the Help panel inside the app, so the two never disagree, and each article ends with a way to act on it: a link into Corpole or a sentence to say to your agent.
Start with your role
- Founder“Is the agent building what I actually want — and will it hold up when others join?”Start here
- Solo builder“Why does my agent forget what we decided yesterday?”Start here
- Team developer“Which work can I pick up, and what does my agent read?”Start here
- Product owner“How do I judge what the agent proposed?”Start here
- Project lead“What may the agent do on its own?”Start here
- Tester“How do I know a criterion is really met?”Start here
- Admin“How do I apply rules and policy to every project?”Start here
- Evaluator“How is this different from Jira plus a CLAUDE.md?”Start here
- Security & compliance“What can an agent reach, and what is logged?”Start here
Browse by topic
- Getting startedFirst steps, the main screens and how a project begins.12 articles
- Connect your agentClaude Code, Cursor and other MCP clients; identity and permissions.9 articles
- Records & authorityRecord types, the four states, writing and approving.19 articles
- Relations & conflictsTyped links between records, conflicts and replacements.3 articles
- Work & boardsThemes to subtasks, Kanban boards and who may move a card.7 articles
- KnowledgeDecisions, questions, glossary, standards, lessons and ideas.5 articles
- Agent policy & rulesWhat an agent may do alone, the rules it follows, notices.9 articles
- Traceability & qualityAcceptance criteria, tests, CI results, defects and risks.8 articles
- Existing codebasesOnboarding code you already have, and handing authority over.3 articles
- When things go wrongSituations and what to do: drift, conflicts, emergencies.20 articles
- Team & accountsMembers, roles, tenants and project settings.7 articles
- Security & dataWhere data lives, agent identity, masking and logs.3 articles
- TroubleshootingError messages and connection problems, with the fix.8 articles
- ReferenceTools, commands, states, relations and terms in one place.7 articles
When things go wrong with your agent
- “My agent forgot what we decided yesterday”A decision made in one chat session is gone in the next. Decisions belong in records, not in chat history.
- “My agent changed a requirement to fit its code”The code did not meet the requirement, so the agent tried to change the requirement instead. In Corpole it can only propose.
- “The agent says "done" but a criterion is not met”The agent reported the work finished, but a criterion has no proof, or its test is failing.
- “The same test keeps failing”The agent is going round in circles, trying variations of the same fix.
- “The agent wants to run a migration or add a dependency”Some actions should never happen without a person. The agent policy decides which.
- “Production is down: declaring an emergency”The fix cannot wait for a requirement. A time-boxed emergency lets the agent fix it now, on the record.
- “Two agents picked the same story”Corpole shows who is on a story and warns a second agent. It does not block; people decide.
- “Our CLAUDE.md is 800 lines and contradicts itself”Move the rules into Corpole as proposals, resolve the contradictions, and hand authority over step by step.
Error messages and quick fixes
- Claude Code shows corpole as failedThe MCP server is listed but not connected. Usually Claude Code has neither a token nor an OAuth sign-in for it.
- 401 invalid_tokenCorpole refused the agent's credentials. The token is revoked or expired, the project is not active, or its owner can no longer read the project.
- My agent ignores the requirementsThe agent writes code without reading Corpole, or follows a file in the repository instead. Most often the repository does not tell it that Corpole comes first.
- The agent's proposal was refusedpropose_record and the other write tools need the propose permission. A read-only connection can read everything its owner can, but cannot propose.
- The GitHub check fails on my pull requestThe Corpole check wants every pull request to cite existing, non-retired record or work item codes.
- "You don't have permission"Corpole refused an action because you, or the person your agent acts for, lack the permission. The message names the permission, who holds it and where it is done.
Still stuck?
Write to info@projera.com with the affected account or project, roughly when it happened and what you saw. Never send passwords, access tokens or payment details; we will never ask for them.