AI Task Orchestration turns scattered requests into a ranked, sequenced backlog. Intake arrives from forms, email, Slack and Microsoft Teams messages, and GitHub or Jira issues. A triage agent classifies each item by type, team and component, attaches a confidence score and proposes a priority and a place in the plan. Nothing changes silently: every reassignment or re-prioritization can pass through an approval gate, and every agent action is written to the audit log. Leads start the day with a clean queue, and requesters can see where their work landed and why.
Every intake event enters the same pipeline. The connector normalizes the payload into a draft task, and the Synapse Engine model router sends classification to a small, low-latency model while longer reasoning goes to a frontier model. Permission-aware retrieval pulls similar past tasks the requester is allowed to see, plus recent human corrections from your workspace as examples, which grounds both classification and duplicate checks. The triage agent holds a scoped, short-lived token that can read tasks and create suggestions through the public Synapse API, but it cannot write to tasks directly. Approved changes are applied under the approver's identity.
The module automates the decisions a lead would otherwise make dozens of times a day: what is this, is it new, how important is it, and when can we do it. Each answer is explainable and reversible, and each can be tuned per team. Changes to scoring weights and escalation rules are recorded in the audit log, so adjusting how a team ranks work is a visible decision rather than a hidden setting.
Orchestration is exposed through the same REST and GraphQL API the triage agent uses. Create tasks from your own intake services with POST /v1/tasks, submit triage suggestions with POST /v1/triage/suggestions, the endpoint the agent itself calls, and subscribe to the task.triaged webhook to sync approved decisions into downstream tools. Suggestions from your code pass through the same approval policies as the agent's. Webhook payloads are signed with HMAC-SHA256 and retried automatically, and the official TypeScript and Python SDKs handle authentication, pagination and signature checks. Requests above your plan rate limit receive a 429 response with a Retry-After header.
| Item | Detail |
|---|---|
| Intake sources | Forms, email, Slack, Microsoft Teams, GitHub, Jira |
| Create tasks | POST /v1/tasks |
| Triage suggestions | POST /v1/triage/suggestions (subject to your approval policy) |
| Webhook | task.triaged (HMAC-SHA256 signed, automatic retries) |
| Active tasks | Up to 25,000 per project |
| API rate limit | Team 300 · Business 1,000 · Enterprise 3,000 req/min (burst 2×) |
Triage runs on the governance layer of the Synapse Engine. Retrieved context is filtered by the requesting user's permissions before it reaches a model, so a suggestion never exposes details from a private space. Intake content is screened for prompt injection, which matters when anyone can file an email or an issue, and approval gates are enforced by the platform API rather than the interface, so a manipulated message has no route around them. Model outputs are validated against a schema before they touch a task. Customer data is not used to train foundation models.
AI Task Orchestration is included in the Team, Business and Enterprise plans. Prices are flat per workspace in EUR, not per user, and each plan is bought as a prepaid term of 1, 3, 6 or 12 months. Team is the entry point at €13 for one month, or €100 for 12 months, the best value, for up to 15 members. Business covers up to 50 members and Enterprise up to 150. Modules are bundled into plans, so there is nothing extra to buy. Starter covers the core workspace only, and upgrades between plans and prorations are handled by our support team.
| Plan | Included | Price (flat per workspace, EUR) |
|---|---|---|
| Starter | No | €10 / 1 month · €25 / 3 months · €45 / 6 months · €70 / 12 months (best value) |
| Team | Yes | €13 / 1 month · €35 / 3 months · €60 / 6 months · €100 / 12 months (best value) |
| Business | Yes | €15 / 1 month · €40 / 3 months · €70 / 6 months · €120 / 12 months (best value) |
| Enterprise | Yes | €20 / 1 month · €50 / 3 months · €85 / 6 months · €150 / 12 months (best value) |
Only where your policy allows it. Low-impact fields such as type can be applied above a confidence threshold you set, while reassignments and priority changes can require approval. Every decision is logged with the actor, policy version and confidence at that moment.
Each task receives a weighted impact × urgency × effort score. The breakdown shows every factor and its weight, and team leads can tune the weights so the ranking reflects how that team actually makes trade-offs.
New items are embedded and compared with open tasks the requester can see. Matches at or above a 0.92 similarity threshold, adjustable per workspace, produce a merge suggestion that a person accepts or dismisses.
No. Without it, auto-planning uses the cycle length and team velocity you configure. With the Planner enabled, it also respects individual capacity, PTO calendars and skills, so proposed sprints do not overload anyone.
Each project supports up to 25,000 active tasks. API usage follows your plan rate limit: 300 requests per minute on Team, 1,000 on Business and 3,000 on Enterprise, with 2× burst and a 429 plus Retry-After when exceeded.
Release notes, engineering deep dives and delivery playbooks from the Synapse team.
© 2026 Synapse AI Systems LLC. All rights reserved.
Third-party product names and logos are the property of their respective owners; their use does not imply endorsement.