Suede AI · Agent directory
Devin agent evidence
A hosted coding agent for repository tasks and pull-request workflows. Review its edition-specific evidence and permissions.
Cognition · Official product · Rating methodology
Ratings measure documentation and evidence coverage. They do not measure agent capability or security.
Cloud
hostedReviewed 2026-10-27 by Codex editorial source review.
Requirements
- Registered Cognition account with cloud-agent access and repository setup. Deployed service revision is not disclosed in this intake.
Permissions and data
- Cloud Agent sessions can modify code, execute commands and browse. Optional profiles restrict access; policy changes may require session resume.
- Publisher discloses default training use with plan-dependent opt-out; review current data controls and terms.
Costs
- Paid cloud-agent access and usage allowances vary by plan; additional usage can incur charges. See dated pricing evidence and current official terms.
Official access / setup · Service terms
Public source: not reviewed. Deployed service revision: not disclosed in this intake.
Performance: Not tested
No task-success, reliability, latency or cost benchmark has been run by this directory.
Security: Not assessed
No independent security assessment has been performed. Documentation points do not establish safety.
No findings recorded in this intake. This does not mean no vulnerabilities exist.
Documentation — 6/10 checks reviewed
| Check | Review | Evidence and reason |
|---|---|---|
| Intended tasks | supported | First-session guide describes code changes, debugging and PR tasks. [source] |
| Boundaries | unknown | Not reviewed in this intake; no point or failing score is inferred. |
| Setup instructions | supported | First-session guide sequences repository setup, Ask and Agent sessions. [source] |
| Prerequisites | supported | Account registration and repository indexing/setup are required. [source] [source] |
| Model requirements | supported | First-session guide explains agent selection and availability. [source] |
| Worked example | unknown | Not reviewed in this intake; no point or failing score is inferred. |
| Tools and integrations | supported | Guide identifies commands and browser use. [source] |
| Permission description | supported | Profiles explain scoped restrictions and when they apply. [source] |
| Failure and recovery | unknown | Not reviewed in this intake; no point or failing score is inferred. |
| Support route | unknown | Not reviewed in this intake; no point or failing score is inferred. |
Hosted disclosures — 6/10 checks reviewed
| Check | Review | Evidence and reason |
|---|---|---|
| Service terms | supported | Official platform terms apply to registered service access. [source] |
| Pricing and usage units | supported | Pricing explains allowances and extra model-dependent usage. [source] |
| Data retention | supported | Publisher states retention policy and exceptions. [source] |
| Training and data use | supported | Publisher states training defaults and plan-dependent opt-out. [source] [source] |
| Authentication | supported | Terms describe account/password registration. [source] |
| Permission controls | supported | Profiles describe optional and mandatory restrictions. [source] |
| Audit and export | unknown | Not reviewed in this intake; no point or failing score is inferred. |
| Deletion and revocation | unknown | Not reviewed in this intake; no point or failing score is inferred. |
| Service status | unknown | Not reviewed in this intake; no point or failing score is inferred. |
| Product changes | unknown | Not reviewed in this intake; no point or failing score is inferred. |
Source receipts
Your First Session: prerequisites, Ask/Agent modes and choosing an agent
Publisher documentation · 2026-10-27T05:02:48Z · Codex editorial source review
- Requires repository indexing/setup before a cloud session. Describes choosing repositories, planning in Ask mode and sending the plan to Agent mode.
- Describes code changes, commands, browser use and pull requests. Explains selectable agents and differing cost/capability; availability depends on administration and plan.
Receipt integrity
Receipt SHA-256
d5720a200ba6c647845e537d1af1607bf1126a64801f10f4679afbdf62f996f1This receipt preserves dated observations. It does not preserve or fingerprint the complete live page or deployed service.
Restrictions; enforcement; changes taking effect
Publisher documentation · 2026-10-27T05:02:48Z · Codex editorial source review
- Describes optional network and MCP allowlists, git read-only/full scope and GitHub CLI token restrictions.
- Recommended profiles may be overridden; mandatory profiles constrain lower scopes. Existing sessions resolve changed restrictions on resume/reboot, not automatically mid-session.
- These documented controls were not tested and are not proof of isolation or safety.
Receipt integrity
Receipt SHA-256
c7e8887b45d12f7cb85506c58fd10bcdd496bdd435388eed0ffa425e9c4affe4This receipt preserves dated observations. It does not preserve or fingerprint the complete live page or deployed service.
June 30, 2026 terms: sections 2, 3.3, 5.1
Publisher documentation · 2026-10-27T05:02:48Z · Codex editorial source review
- Public terms govern registered access to the Cognition platform and describe account/password responsibilities and usage restrictions.
- Terms allow customer data use for training by default, with a paid-tier opt-out and Teams administrator control; exceptions to zero retention include abuse and legal process.
Receipt integrity
Receipt SHA-256
51343505259e49c506b9d837c8b0b1922f597621611c5408b4c56692938e383cThis receipt preserves dated observations. It does not preserve or fingerprint the complete live page or deployed service.
Privacy & Intellectual Property: retention and model training
Publisher documentation · 2026-10-27T05:02:48Z · Codex editorial source review
- Publisher states customer data is retained for the customer relationship unless otherwise specified; feedback and interaction data may be kept as long as Cognition determines necessary.
- Publisher states training is enabled by default with paid-plan opt-out; enterprise training requires express written consent. This receipt records disclosures, not compliance verification.
Receipt integrity
Receipt SHA-256
41f2240322fb979b8114ed1a1b899c9d652d3ad703998d8e6a5ce9b475ee2c3aThis receipt preserves dated observations. It does not preserve or fingerprint the complete live page or deployed service.
Paid plans and usage FAQ
Publisher documentation · 2026-10-27T05:02:48Z · Codex editorial source review
- Pricing page lists paid access to cloud agents and explains daily/weekly included allowances and additional usage at API pricing.
- Per-message cost varies with model, task size and reasoning. No purchase or billed API call was made; check current pricing before use.
Receipt integrity
Receipt SHA-256
254682f95490d9c160dbdcab14b98701b41774bc81e0b57cee15d079441cd220This receipt preserves dated observations. It does not preserve or fingerprint the complete live page or deployed service.
Download full report · Report version 258801dab7fb · Request a recheck or correction