Bring any agent. Yours stays yours.
Every project tool now has agents working tasks. The differences that matter are underneath: whose model runs the work, which agents are allowed in, what stops a bad write, who can prove what happened, and what the bill looks like. Planra is the only one where you bring your own keys and your own agents — with approval gates, provenance, and per-agent budgets enforced on every write, including MCP.
No seats. No credit anxiety. Pay for work done.
| Product | BYO model (your API keys) | BYO agent (any external agent) | Pre-action approval gates | Agent provenance | Per-agent budget gating | Pricing |
|---|---|---|---|---|---|---|
| Planra | YESAny key: OpenAI, Anthropic, Gemini, custom endpoints | YESAny agent via MCP — trust-gated, not just CRUD | YESRisk-tiered: destructive/outbound/costly, per-agent trust dial | YESFull timeline + compliance export, human vs agent on every action | YESPer-agent caps that pause work — visibly, with alerts at 50/80% | Pay for work done |
| ClickUp | NO | NO | YESApproval Mode (all-or-nothing) | PARTIALFull audit log is Enterprise-only, 30-day window | PARTIALCost panels — transparency, no gating | Seats + credits |
| Jira (Atlassian) | NO | PARTIALCurated agent catalog | PARTIALInherited from workflow approvals | YES | NOCredits are billing, not governance | Bundled + credits |
| monday.com | NOVendor-managed models | YESMCP + managed-agent import | PARTIALPermission-scoped | YES | PARTIALPer-user credit limits, not per-agent | Seats + mandatory credits |
| Notion | NOModel picker, billed via credits | YESExternal Agents API (partner roster) | PARTIALPlan mode on personal agent only | YES | PARTIALWorkspace pool auto-pause, not per-agent | Seats + credits |
| Asana | NO | NO | PARTIAL | PARTIAL | PARTIALDepartment credit allocations, soft limits | Seats + credits |
| Linear | NO | YESFixed third-party roster | NOGuidelines, not enforcement | NOAudit log covers account events only | NO | Seats + AI credits |
| Wrike | NOAzure-OpenAI locked | PARTIALInbound MCP under user RBAC | NOAgents can start approval flows; no per-action gate | YES | NO | Seats + action packs |
| Motion | NO | NO | PARTIALManual mode or fire-and-forget | NO | NO | Seats + AI Employee tiers |
| Taskade | NOBYOK is an open feature request | PARTIALMCP connectors | YESPer-tool approval toggle | NOUsage dashboard only | NO | Seats + credits |
| Plane | PARTIALCloud is Plane AI credits only; self-hosted Commercial Edition can run its own AI services | YESMCP server + agent framework; agents can be work-item assignees | PARTIALBuild mode reviews action cards, Autopilot skips review; per-transition approvals are Enterprise | PARTIALActivity trail + Agent Run lifecycle; audit-log API is Enterprise-only | PARTIALPer-seat credit pools with admin limits on Business — not per-agent caps | $6–13/seat + credits; free ≤12 users; AGPL self-host |
| Dart | NOVendor-hosted model picker | PARTIALMCP read/write with identity token | NO | PARTIALMCP write attribution only | NO | Seats, AI bundled |
Compiled July 2026 from vendor documentation, release notes, and public community threads; the Plane row was refreshed in September 2026 against plane.so, docs.plane.so and developers.plane.so. “NO” means absent from public documentation at time of writing — vendors ship fast; tell us if something changed and we’ll update it.
Export a CSV (or a Trello JSON) and drop it on Workspace → Import. Planra detects the tool from the file, then brings columns, tasks, statuses, priorities, labels, estimates and dates across in one step. Leaving is just as easy: every member can download the whole workspace as JSON from Settings, and provenance exports to CSV. Your work never needs a migration service.