Documentation
Glossary
The terms used across the eClips Workforce docs. Underlined words in the manuals link here.
- eClips
- The company. eClips Tech builds several products, and each has its own name. When you mean the AI operations product, say "eClips Workforce", not just "eClips".
- eClipser
- Someone who has completed an eClips Academy track for a product: Workforce User, Administrator or Developer.
- Agent
- An AI worker that does one job, such as processing invoices or reviewing contracts. eClips Workforce has 102 agent types across finance, HR, procurement, legal, operations, sales, IT and security, marketing and customer service.
- Agent run
- One execution of an agent on one input. Every run is stored with a status, a confidence score, the model used, and the tokens and credits it consumed.
- Triage
- The exception queue. When an agent isn't confident enough, or an action needs sign-off, the item goes to the Approval Inbox for a person to approve or reject.
- Confidence score
- A score from 0 to 1 that an agent gives its own output. At 0.85 or above a run can go ahead. Below 0.85 it goes to triage, and below 0.50 the triage reason says the score was under the minimum.
- Harness
- The runtime wrapper around every agent run. It handles retries with backoff, timeouts, the circuit breaker, model escalation and downshift, the evaluator on high-stakes agents, and the credit meter.
- Model downshift
- When a plan doesn't include a premium model, the harness runs the agent on the best model the plan allows instead of blocking it. Every plan can use every agent.
- Model escalation
- When a fast-model run comes back with low confidence, and the plan allows it, the harness retries once on Claude Opus and keeps the new result only if confidence improved.
- Evaluator
- A second agent that grades the output of six high-stakes agents (such as payroll, tax and fraud). It can send the run to triage or ask for an Opus retry. It can never approve something on its own.
- Trust level
- How much an organization lets an agent type do on its own, from 0 (observe only) to 4 (full autonomy). An admin sets it in Settings, and every agent starts at 0. Separately, each agent builds an earned autonomy stage from its track record; see Trust & Governance in the System Manual.
- Context Book
- What your agents know about your organization: company details, key people, the approval matrix, custom rules and vendors they must never deal with. You fill it in during onboarding and edit it in Settings.
- Credit
- The unit of usage. One credit is 1,000 tokens on a fast model. Heavier models use more credits per token. Each plan includes a monthly credit allowance.
- Credit multiplier
- The factor applied to token usage for each model tier: fast (Gemini Flash, Claude Haiku) 1x, smart (Claude Sonnet) 12x, premium (Claude Opus) 60x.
- Member budget
- A monthly credit cap an admin sets for one person, so nobody can use up the shared pool. Available on Team, Growth and Enterprise plans.
- Scope
- A permission attached to an API key, such as agents:run, triage:write or webhooks:read, that limits what the key can do.
- API key
- A secret token, prefixed eck_live_ or eck_test_, that authenticates REST API and SDK requests. The key identifies your organization, so you never send an organization id.
- Reach
- The browser tool in eClips Workforce that agents use when a system has no API. It works the web interface directly, and every submit or destructive step waits for a person to approve it.
- MCP
- Model Context Protocol. The connector layer that lets agents call outside systems such as an ERP, a CRM or an accounting package through one interface.
- Organization
- A tenant in eClips Workforce. Its data, credits, members and settings are kept apart from every other organization by row-level security.
- Workspace
- Another word for organization: your team's own environment, identified by a unique slug.