01
Sequential Execution
Stages execute in a defined order rather than running as disconnected tasks.
← DocumentationDESIGN.
Agent Flow Guide
DESIGN.
CONNECT.
REPEAT.
Agent Flows connect your recruited specialists into a reusable sequential workflow. Each stage contributes its own specialty before AgentForge creates one coordinated final result.
2+
Workflow Stages
10 / 30 CR
Per Stage
1
Final Synthesis
Saved Workflows
OUTPUT BECOMES
THE NEXT INPUT.
Agent Flows are designed for work that benefits from a specific order of specialist execution.
Instead of asking several agents to independently answer the same mission, each stage receives useful findings produced before it. This lets research, development, content, QA, prompt design, automation, visual creation, security review and mission coordination build on one another in a deliberate order.
Example Workflow
Lyra→Nexus→Vex→Rune→Final Synthesis
Example: Lyra researches the problem, Nexus turns findings into a technical solution, Vex validates quality, and Rune converts the final requirements into a production-ready prompt. Any recruited AgentForge specialists can be arranged into a Flow.
Workflow Lifecycle
BUILD THE
PIPELINE.
01
Create a Flow
Create a reusable AgentForge workflow and define the purpose of the process you want to repeat.
02
Add Specialist Stages
Assign recruited specialists to ordered stages and optionally add stage-specific instructions.
03
Plan Execution
AgentForge evaluates each stage independently and determines whether it should run in Standard or Research mode.
04
Pass Output Forward
Completed specialist output becomes context for the next stage so the workflow builds progressively.
05
Synthesize the Workflow
After all specialist stages finish, AgentForge combines their work into one polished final deliverable.
Flow Architecture
BUILT FOR
REPEATABLE AI WORK.
02
Stage-Level Pricing
Each stage receives its own execution mode and credit cost before the mission begins.
03
Persistent Stage Results
Completed stage outputs are stored so recoverable failures do not require every specialist to run again.
04
Configuration Integrity
Started missions preserve the execution configuration they launched with so later edits do not silently change historical work.
Flow Credits
EACH STAGE
HAS A COST.
AgentForge evaluates every workflow stage independently. Standard stages cost 10 credits while stages that require research cost 30 credits.
Standard + Standard
20 CR
Research + Standard
40 CR
Research + Research
60 CR
Resumable Execution
DON'T REPEAT
COMPLETED WORK.
AgentForge stores the state of each workflow stage. If a later stage fails before billing completes, already-completed stages can be loaded from saved execution state instead of calling the AI again.
Completed
Reuse saved output
Failed
Retry required stage
Billing Retry
Preserve final result
Already Charged
Finalize only — no rerun
Configuration Integrity
HISTORICAL RUNS
STAY CONSISTENT.
Once execution begins, AgentForge preserves the stage configuration used by that mission. Editing the original saved Flow later should affect future missions rather than silently changing the meaning or pricing of an execution already in progress.
Execution Snapshot
Agent
Stage Instruction
Execution Mode
Credit Cost
Execution Protection
CHECKED AT
EVERY STAGE.
01
The flow must belong to the authenticated user.
02
The saved flow must still be active before a new execution begins.
03
Every participating specialist must still be recruited and usable under the current plan.
04
AgentForge calculates the complete flow cost before claiming the mission.
05
Completed specialist stages can be reused during a safe retry.
06
Generated final results are preserved during recoverable billing failures.
07
Existing credit transactions prevent duplicate charging.
Next Guide
UNDERSTAND YOUR CREDITS.
Learn how AgentForge tracks balances, mission charges, execution history and plan allowances.